Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
Research ArticleSenStick Comprehensive Sensing Platform with an Ultra TinyAll-In-One Sensor Board for IoT Research
Yugo Nakamura Yutaka Arakawa Takuya KanehiraMasashi Fujiwara and Keiichi Yasumoto
Graduate School of Information Science Nara Institute of Science and Technology 8916-5 Takayama-cho Ikoma Nara 630-0192 Japan
Correspondence should be addressed to Yugo Nakamura nakamurayugons0isnaistjp
Received 13 June 2017 Revised 31 August 2017 Accepted 20 September 2017 Published 26 October 2017
Academic Editor Alberto J Palma
Copyright copy 2017 Yugo Nakamura et alThis is an open access article distributed under theCreativeCommonsAttribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited
We propose a comprehensive sensing platform called SenStick which is composed of hardware (ultra tiny all-in-one sensor board)software (iOS Android and PC) and 3D case data The platform aims to allow all the researchers to start IoT research such asactivity recognition and context estimation easily and efficiently The most important contribution is the hardware that we havedesigned Various sensors often used for research are embedded in an ultra tiny board with the size of 50mm (119882) times 10mm (119867) times5mm (119863) and weight around 3 g including a battery Concretely the following sensors are embedded on this board accelerationgyro magnetic light UV temperature humidity and pressure In addition this board has BLE (Bluetooth low energy) connectivityand capability of a rechargeable battery By using 110mAh battery it can run more than 15 hours The most different point fromother similar boards is that our board has a large flash memory for logging all the data without a smartphone By using SenStickall the users can collect various data easily and focus on IoT data analytics In this paper we introduce SenStick platform and somecase studies Through the user study we confirmed the usefulness of our proposed platform
1 Introduction
The rapid growth of Internet of Things (IoT) has made itpossible to sense many types of information from the realworld In the research area of human activity recognition IoTdevices equipped with various sensors have been widely usedas a tool for sensing the activities There are three approachesof activity recognition studies in terms of sensor deployment
The first approach deploys sensors into the environmentlike a smart home As typical sensors camera [1] sound [2]and power consumption [3] have been used for monitoringdwellerrsquos activities
In the second approach sensors are attached to thehuman The smartphone has been widely used [4 5] as asensing tool in this approach because it has many sensors(an acceleration sensor gyro sensor magnetic sensor etc) aswell as a processor a large data storage and wireless networkNowadays instead of the smartphone wearable devices suchas a smartwatch are used [6] JINS MEME [7] is a smarteyewear having the electrooculography (EOG) electrodesin the bridge of the glasses and the nose pads It tries to
measure the movement of eyes for recognizing the internalcontexts such as drowsiness and tiredness Also we haveproposed Waiston Belt [8] that can measure the abdominalcircumference and posture
The third approach is a new paradigm where sensors areembedded in the various things surrounding us Kurahashiet al [9] have installed an acceleration sensor into a toiletpaper holder for recognizing the person HAPIfork [10] hasembedded an acceleration sensor into a fork for monitoringeating activities We have also added an acceleration sensoron a faucet of water or a remote control for sensing variousactivities of the dwellerWe think that the spread of IoTmakesmore things have a sensor and network connectivity In thefollowing sections we define thewords ldquosensorize (verb)rdquo andldquosensorization (noun)rdquo as the procedure to add the sensor andwireless network to the nonelectronic personal belongings
However sensorization of tiny things is difficult andcostly for the researcher Nowadays various rapid proto-typing platforms such as Arduino Mbed and Raspberry Pihave already spread by the maker movement They are cheapand useful but there are some problems for sensorizing tiny
HindawiJournal of SensorsVolume 2017 Article ID 6308302 16 pageshttpsdoiorg10115520176308302
2 Journal of Sensors
Hardware
SenStick
Soware
3D data
Sensorizeevery thing
+
IoTeducation
+ Create a community3D case data
+ Share 3D data
Multiplatform(iOSAndroidPC)+ Functions for research+ Functions for education
SenStick 2+ Shorter+ Separable+ Wireless sync+ Long life
Wireless charging model+ inner+ Wireless charging+ Sealed case
Research on IoT IoT education
SenStick 1+ Ultra tiny (3 g)+ Multisensors+ BLE+ Memory+ Battery 2017
SenStick 3+ Commercialized+ nRF52 (Cortex-M4F)
Open Source
2015
2016
Figure 1 SenStick platform including hardware software and community
thingsmentioned as followsThe size of the board is relativelylarge In addition the final deliverables usually become largerbecause the baseboard requires various additional functionssuch as a charging circuit clock and storage and those areconnected via the relatively large connector As a result itis actually difficult to develop ldquosmallrdquo sensorized things thatcan be used in the actual environment Another problem issoftware for controlling the deliverable from a smartphone Itis not easy to develop a sophisticated mobile application foreach deliverable respectively Since the essence of IoT is dataanalysis sensorization of everything should be simplified andsolution for collecting data smoothly is required
Based on above-mentioned background we starteddeveloping a platform called SenStick [11] for sensorizingvarious things more easily in 2014 Our proposed SenStickis a comprehensive platform including the hardware soft-ware 3D case data and community as shown in Figure 1SenStick aims to allow researchers to sensorize everythingand also educators to teach IoT-related techniques such as
data analysis The size of the first SenStick board releasedin 2015 (SenStick 1) is 75mm (119882) times 10mm (119867) times 5mm(119863) and its weight is around 3 g including a battery On thistiny board 8 sensors (acceleration gyro magnetic light UVtemperature humidity and pressure) flash memory BLEbattery and charging circuit are densely embedded
The second version of the hardware (called SenStick 2)released in 2016 [12] became shorter 50mm (119882) times 10mm(119867) times 5mm (119863) by dividing the board into two boards andconnecting them vertically By updating the firmware thebattery life of SenStick 2 reached 15 hours by 60mAh batteryAlso wireless data synchronization function and DeviceFirmware Update (DFU) over BLE have been implementedSenStick 2 has a derivative version (wireless chargingmodel)This version has a different battery type and charging circuitfor enabling wireless charging
In 2017 the third version of the hardware (called SenStick3) was sold from a company (httpsenstickcom) Almostall the configurations of SenStick 3 are the same as SenStick
Journal of Sensors 3
2 but the board is unified into one board again like SenStick1 The main change is the upgrade of BLE chip from nRF51 tonRF52 Since the processing power is increased we can addsome calculation process on the board
In the software part support application is developedfor mobile devices (iOS and Android) and desktop PC(LinuxmacOS andWindows) Each application canmonitorand record sensing data synchronously and can configurethe parameters of each sensor The Android version cansimultaneously record a ground truth video and sensing dataThe PC version allows a user to connect multiple SenStickssimultaneously and also provides a function to use Node-RED which provides a simple programming interface
In 3D case part we designed the data of basic 3D case ofSenStick It is possible to easily design the 3D case for variousthings by using our base 3D case data
The remaining part of the paper is organized as followsWe first introduce various existing products and comparethem with SenStick in Section 2 Next we explain the detailsof our proposed platform SenStick in Section 3 In Section 4we introduce some case studies utilizing SenStick and somesurvey results answered by students who actually used Sen-Stick platform Finally we conclude the paper in Section 5
2 Related Work
We summarize competitive sensors in the market in Table 1SensorTag [13] of Texas Instruments is the most competi-
tive sensor board similar to SenStick The sensors embeddedon SensorTag are almost the same as SenStick except for UVsensor However SensorTag does not have stand-alone datarecording function and a flash memory because this sensoris designed on the premise that sensing data are always storedon the smartphone connected by BLE Since the main targetof SensorTag is environmental sensing the sensing frequencyof motion sensors is limited up to 1Hz and a rechargeablebattery is adopted Therefore it is not suitable for activitysensing of IoT research In addition the board shape size andweight are completely different SenStick is much smaller andslimmer
IoT Smart Module [14] is a small sensor board equippedwith several sensors like SensorTag Support applicationsprovide functions to cooperate with IBM Bluemix so itis possible to smoothly perform tasks from sensing ofdata to cloud analysis However support applications onlywork on Android Similar to SensorTag of Ti it adoptsnonrechargeable coin battery Although it can set highersensing frequency a user must change the coin battery atshort intervals Also logically it cannot transmit all thesensing data in case of higher sensing frequency becausethe transmission capacity of BLE is low Therefore it is notsuitable for research that requires precise data
TSND121 [15] of ATR-Promotions is another competitivesensor because it was used in some academic research Ithas 9-axis motion sensors and ambient pressure sensor andadopts Classic Bluetooth 20 for transmitting data Becauseit only focuses on monitoring activities of a human and arobot the size of the sensor is relatively large but at most7 sensors can be simultaneously connected to a control PC
Since it always requires a PC for recording sensed data it isnot suitable for sensing our daily life out of the lab
WAX9 [16] of Axivity used in UK Biobank [17] was astreaming inertialmeasurement unit (IMU)which has a largeflash memory for logging The sensor combines a MEMSaccelerometer gyro and magnetometer with a Bluetoothlow energy (BLE) compatible radio Additionally WAX9 alsofeatures a barometric pressure sensor and temperature sensorHowever it does not have some environmental sensors suchas humidity light and UV sensor and always only distributesthe Windows-compatible software for data collection andvisualization It has already been discontinued
Based on the features of other competing sensors wesummarize the requirements for SenStick platform
(1) Hardware Requirements The SenStick platform aims tosense tiny activities that cannot be sensed by existing smart-phones and wearable devices Therefore essential functionsmust be embedded in a small sensor board and each sensormust collect data in high sampling rate Since BLE commu-nication dose not guarantee reliable data communicationit is necessary to be able to record in standalone withoutrecording devices such as a smartphone In addition it isrequired that the SenStick must have rechargeable batterysystem because a coin battery is not suitable for frequent use
(2) Software Requirements The SenStick platform aims tosimplify the data collection and analysis from sensorizedthings Therefore support software should be multiplatformIt should be compatible with the various operating systemsSupport software requires not only basic function such as datarecording but also additional functions such as simultaneousrecording of sensing data frommultiple SenSticks and simpleprogramming interface for data processing
(3) 3D Case Requirements The SenStick platform aims topromote the sensorization of everything For this purposeit is desirable to provide a 3D case of SenStick that can beattached to various belongings Further 3D cases createdby someone should be shared with other users through ourwebsite
3 Comprehensive Sensing Platform forIoT Research SenStick
Our proposed platform called SenStick consists of 3 com-ponents hardware software and 3D case In the followingsubsections we explain the details of these components Thisplatform allows a user to sensorize everything easily andefficiently We believe that it supports the rapid prototypingby researchers and is useful for IoT education (a videoillustrating SenStickrsquos hardware software and 3D case datais provided in Supplementary Material available online athttpsdoiorg10115520176308302)
31 Hardware Aswe explained in Section 1 we have updatedthe hardware three times from 2015 The explanation in thefollowing part is based on SenStick 2 developed in 2016The size of the sensor board is 55mm (119882) times 10mm (119867)
4 Journal of Sensors
Table1Com
paris
onof
competitives
ensors
SenS
tick
3Se
nStic
k 2
Sens
orTa
g (C
C265
0)Io
T Sm
art M
odul
eTS
ND
121
WA
X9
Size
(mm)
50(119882
)times10
(119867)times
10(119863
)50
(119882)times
10(119867
)times10
(119863)
42(119882
)times32
(119867)times
10(119863
)44
(119882)times
27(119867
)times11(119863
)37
(119882)times
46(119867
)times12
(119863)23
(119882)times
325(119867
)times76
(119863)
Weight(g)
35
28mdash
227
Stand-alon
erecording
Itimes
timestimes
ISm
artpho
neapp
II
Itimes
timesDesktop
app
I998779
timesI
IBa
ttery
life
15(hou
r)1(year)
1(year)
6(hou
r)6(hou
r)Ba
ttery
type
Rechargeablerem
ovable
Rechargeable
Coinbatte
ryCoinbatte
ryRe
chargeable
Rechargeable
Network
BLE
BLE
BLE
Bluetooth20
BLE
Datap
rocessing
Itimes
timestimes
timestimes
I2CEx
tension
Itimes
timestimes
timestimes
Sensors
Acceleratio
nI
II
II
Gyro
II
timesI
IMagnetic
II
II
IAirpressure
II
II
ITemperature
II
Itimes
IHum
idity
II
Itimes
timesLight
II
Itimes
timesUV
Itimes
Itimes
timesMi c
timesI
timestimes
times
Journal of Sensors 5
Flash memory(back side)
50GG
60GB
Temperature amphumidity(SHT-20)
UV(VEML6070)
Pressure(LPS25HB)
Light(BH1780GLI)
Charger(BQ24090)
Micro USB
Li-Po battery
USBUART(CP2104)
9-axis(MPU-9250)
BLE(nRF51822)
Figure 2 Hardware of SenStick 2 (8 sensors BLE flash memory etc)
Table 2 Sensors embedded in SenStick 23
Type of sensor Model number Power consumption9-Axis MPU-9250 280 120583Asim37mAPressure LPS25HBTR 4 120583Asim25 120583ATemperaturehumidity SHT20 270 120583Asim330 120583AIlluminance BH1780GLI-E2 120120583Asim200 120583AUltraviolet (UV) VEML6070 100 120583Asim250 120583A
times 5mm (119863) and its weight is around 3 (g) including thebattery On this tiny board eight sensors (acceleration gyromagnetic light UV temperature humidity and pressure)flash memory (32Mbytes) BLE battery and charging circuitare densely embedded as shown in Figure 2 The modelnumber and power consumption of the sensors embedded inSenStick are listed in Table 2 All the sensors are popular chipsused worldwide and their performances are ensured by eachmanufacturer
The novelty of our hardware is the design of the sensorboard composed of the elaborated combination to satisfythe requirements presented in Section 2 This contributionbreaks down into the following aspects
(1) Combination of BLE and Flash Memory Almost allthe BLE-based devices equipped no large flash memoryto record sensing data because these devices are designedon the premise that sensing data are always recorded onsmartphones connected via BLE However the transmissionspeed of BLE is not enough for sending all the sensing dataof multiple sensors with the high sampling frequency As aresult some data will be dropped in the transmission Sincethe design of existing BLE-based devices is not appropriatefor research purpose we designed a new BLE-based sensingboard with a large flash memory which allows users torecord all the sensing data precisely Also it means that it canwork without holding a smartphone This ability is useful forsensing themotion of animals that cannot have a smartphone
(2) Selection of Powerful and Energy-Saving SoC Weselected Nordicrsquos nRF52832 SoC as SenStickrsquos microcon-troller Because the nRF52832 SoC is built around a 32-bit ARM Cortex-M4F CPU with 512 kB + 64 kB RAM itis extremely powerful Moreover the embedded 24GHztransceiver supports Bluetooth low energy and has extremelylow power consumption and it is the worldrsquos smallest classmodule Each sensor is connected via two I2C buses inde-pendently processing high-speed sensor (acceleration gyroandmagnetic) and low-speed sensor (light UV temperaturehumidity and pressure) taking into account the differencein bus occupancy time This makes it possible to realize highsampling with the motion sensors
(3) Application-Friendly Battery Type and Shape Since Sen-Stick is designed to bemounted on objects such as chopstickstoothbrushes and glasses it is desirable that the shape is thinand compact However sensors such as SensorTag equippedwith readily available coin batteries are inappropriate inshape because the footprint is square Also in the researchnonrechargeable coin batteries are costly and troublesomeFor those reasons we selected a thin and compact lithiumpolymer battery on SenStick This type of battery was harderto obtain than expected Therefore the procurement ofbattery has become the bottleneck of development
SenStickrsquos firmware is newly designed and developed byconsidering aforementioned requirements Generic Attribute(GATT) profile of SenStick provided by firmware is shownin Figure 3 The GATT server in SenStick provides threeservices (SenStick Control Service Metadata Read Serviceand Sensor Service) The SenStick Control Service providesoperation instructions such as starting and stopping sensingand logging The metadata reading service provides thefunction of reading metainformation of log data The SensorService provides functions to specify the sensing and loggingoperation of each sensor (0 acceleration 1 gyro 2 magnetic3 light 4 UV 5 temperaturehumidity 6 pressure) Also itprovides real-time sensing data reading function and log datareading function
6 Journal of Sensors
(32-)
Peripheral
GATT server
SenStick Control Service
Metadata Read Service
Read-write
Data notication
Flashmemory
(0 acceleration 1 gyro 2 magnetic 3 light
4 UV 5 temperaturehumidity 6 pressure)
GATT client
Central
~Sensor Service (0 6)
Figure 3 Generic Attribute (GATT) profile of SenStick
Table 3 BLE communication specification
Type of content SpecAdvertising packet Complete UUID flagScan response Local name (SenStick)Advertising packet interval 100ms (30 s)rarr1000ms (150 s)rarrsleepConnection interval max 80msmin 20msSupervision timeout 4000msSlave latency 0
Basically all the sensing data will be notified to a smart-phone in real time up to a transmission capacity of BLE Ifthe amount of sensing data exceeds the transmission capacitya part of sensing data will be dropped Therefore the real-time sensing data reading function is used formonitoring thestatus or for an interactive application that does not requireprecise data For precise data acquisition the log data readingfunction which guarantees the data reading order is usedThe communication specifications of BLE are described inTable 3
As shown in Table 4 the fixed capacities of a flashmemory are allocated to each sensorThe number of samplesindicates the total number of samples that can be recorded bythe allocated capacities
SenStick stops logging when the number of samples ofone sensor exceeds the storage capacity allocated for it Themaximum recording time of acceleration gyroscope andmagnetic sensor is 4 hours 40 minutes at sampling intervalof 10 milliseconds (sampling frequency 100Hz) and 14 hours10 minutes at sampling period of 30 milliseconds (samplingfrequency 33Hz) The maximum recording time of othersensors is 9 hours 26 minutes at sampling period of 200milliseconds (sampling frequency 5Hz)
32 Software We provide mobile applications for iOS andAndroid and a library for nodejs platform Common func-tions of them are operating function parameter settingfunction and real-time data view function An additional
Table 4 Allocated storage capacity per sensor
Amount of allocatedstorage capacity Samples
Acceleration 102Mbytes 17M samplesGyro 102Mbytes 17M samplesMagnetic 102Mbytes 17M samplesIlluminance 340KB 170K samplesUV (UV) 340KB 170K samplesHumidity 170KB 170K samplesAir temperature 170KB 170K samples
function of iOS application is a firmware update functionThose of Android application and nodejs applications areground truth video recording function and an ability to con-nect multiple SenSticks respectively Since all these softwareapplications are shared on GitHub as an own source softwareall the users can create their application based on their ownpurposes
Communication Protocol Here we explain the communica-tion protocol between SenStick and softwareThe explanationis based on ldquonode-SenStickrdquo which is our developed libraryfor nodejs Figure 4 shows a protocol sequence betweenSenStick and the central terminal (node-SenStick on Mac)
(I) Discovery First of all the application needs to find thesurrounding SenStick by using SenStickdiscovery() methodin our library When SenStickdiscovery() method finds aSenStick a callback function with an instance of SenStickobject containing the SenStickrsquos UUID and LocalName asarguments is called
(II) Connect If the target SenStick is found by the discoveryprocess the application is going to establish a connec-tion with it by calling SenStickprototypeconnectAndSetUp()method After establishing the connection this methodacquires the information necessary for operation from thetarget SenStick
Journal of Sensors 7
Central
Node-SenStick SenStick
SenStickdiscovery()
SenStickprototypeconnectAndSetUp()
Establish connection
SenStickprototypewriteSensorMeasurementCong()
Reect setting
SenStickprototypestartSensingAndLogging()
SenStickprototypenotifySensor()Start sensing
Select sensor to notify
Sensing data
SenStickprototypestopSensingAndLogging()
Stop sensing
Enable log notication
Select log data to notify
SenStickprototypewriteSensorLogReadoutTargetID()
Logging data
SenStickprototypedisconnect()
Disconnect
SenStickprototypenotifySensorLogData()
Periperal
(I) Discovery
(II) Connect
(III) Setting
(V) Obtaining log
(VI) Disconnect
(IV) Start sensingamp
RT receiving
Figure 4 Communication sequence between SenStick and node-SenStick
(III) Setting After establishing the connection the appli-cation sets three parameters (operation mode samplinginterval and measurement range) for each sensor by usingSenStickprototypewriteSensorMeasurementConfig()methodThe operation mode is a parameter that sets valida-tioninvalidation of sensing and logging of each sensor Thesampling interval and measurement range are parameters ofeach sensorwhichwill be changed by the sensing purpose andapplications
(IV) Start Sensing and Real-Time Data Receiving When Sen-StickprototypestartSensingAndLogging() method is calledSenStick starts sensing and logging Only valid sensorsrsquodata are transmitted via BLE and also recorded into aflash memory ID of each log is automatically assignedDuring sensing SenStick notifies sensing data in realtime To receive this notification the application callsSenStickprototypenotifySensor() method (Sensor representseach sensor name) and activates the notification for
8 Journal of Sensors
(i) Frequency(ii) Range
Device select Main view Settings Firmware update
Real-time view Log list in SenStick Download from SenStick
Selected sensorSensor name
Sensing frequency
Range of y-axis
Figure 5 iOS application for SenStick
the sensor When a notification of sensing data isreceived sensorChange event will be issued If Sen-StickprototypestopSensingAndLogging() method is calledSenStick stops sensing and logging
(V) Log Data Reading To read the log data recordedin the flash memory to the application the applicationcalls SenStickprototypenotifySensorLogData()method (Sen-sor represents each sensor name) and enables log datanotification If the application specifies a valid log ID bycalling SenStickprototypewriteSensorLogReadoutTargetID()method SenStick starts informing the sensing data recordedin the log sequentially To receive notification of log datasensorLogDataReceived event will be issued By receiving thisnotification by the application it is possible to acquire the logdata recorded in the memory
(VI) Disconnect After finishing the data acquisition Sen-Stickprototypedisconnect() method will be called to releasethe connection of SenStick And SenStick will begin advertis-ing again
User InterfaceWe explain the user interface of the applicationbased on iOS version Figure 5 shows a sophisticated mobileapplication for SenStick As shown in Figure 5 the userselects target SenStick from devices listThen the applicationshows the main view In the main view the user can selectsensors to be acquired For immediate data monitoring allthe received data via BLE can be shown in the graph view inreal time The application can control all the settings of eachsensor such as sample frequency and sensing range And alsoit has firmware update function Sensing data stored in theflash memory of SenStick is viewed from log list and it can be
downloaded to the smartphone in CSV formatThis functionprevents data loss caused by BLE communication
33 3D Case Data In order to make sensorized things anattachment case for mounting the SenStick is required Wedesigned the basic 3D case of SenStick By extending thisbasic case users can design derivative cases for chopsticksglasses and so on Actually our laboratory has succeeded indesigning cases for several things as shown in Figure 6 Allof these cases data are shared on our community site Eachcase is designed in a few minutes thanks to the basic caseFor example SenStick case for glasses is made by combiningbasic case of SenStick and open 3D data of glasses distributedby i Design Studio In addition due to the popularizationof 3D printers in the world and enhancement of tools andreferences for 3D modeling all the users can print out theirdesired SenStick cases by themselves and attach our sensor toeverything
4 Case Studies of SenStick
SenStick an ultra small sensor can be embedded into varioustiny things in our daily life Our first target is our dailybelongings such as glasses chopsticks and pens becausesensorization of daily belongings is good to intuitively under-standwhat can be sensed andwhich sensor is usefuleffectiveSenStick can upgrade normal glasses to intelligent oneswhichcan monitor a head movement posture and the amount ofsunlight received If SenStick is embedded into chopsticksand pens we can observe eating activities and handwritingactivities respectively In the future we will embed it into astick for elderlymonitoring sports gearsmusical instrument
Journal of Sensors 9
Chopsticks Tooth brush Glasses Umbrella Name card Cookware
A user can design original case data in a few minutes
Basic 3D case of SenStick
Figure 6 3D case data of SenStick
Sensorized
Chopsticks
Figure 7 Eating activity recognition by sensorized chopsticks
kitchen utensils animals stuffed toy a remote control andany kind of things that can be sensing targets In this sectionwe introduce some case studies results utilizing SenStick
41 Eating Behavior Recognition by Sensorized Chopsticks Inmany cases East Asian people eat meals with chopsticksBy analyzing the data obtained from chopsticks there is apossibility that we can estimate the eating habits includinganswers for questions like ldquohow did you eatrdquo and ldquowhatdid you eatrdquo and so forth These data are very useful formaking healthcare applications and leading a healthy lifeThis is one reason why we develop the sensorized chopsticksusing SenStick shown in Figure 7
Research goals of the eating behavior recognition by sen-sorized chopsticks are as follows G1 (behavior) recognizingthe timing and speed and detecting ldquocroprdquo and ldquointakerdquoaction G2 (food type) classifying food content like ricenoodles and so forth from the eating behavior and G3(person) identifying an individual from using chopsticks Inthis paper we show the research results on G1 and G2
We conducted basic experiments to recognize the eatingbehavior (G1 and G2) We used acceleration and gyroscopedata measured by SenStick where the sensing rate is 10HzThe purpose of experiment 1 is to understand the basic
characteristics of sensing data from sensorized chopstickwhen a participant ate chips The purpose of experiment 2 isto identify the food type from the difference in movement ofchopsticks We assumed 4 classes (C1 rice and dish C2 misosoup C3 noodles and C4 shrimp) as food types
Figure 8 shows time series of sensing data for a set ofeating actions in experiment 1 The eating actions are cropaction and intake action In crop action 6-axis motionsensors detect movement (open and close) of the chopsticksIn intake action sensors detect movement of the arm fromthe dish to the mouth Each graph in Figure 8 shows certaincorrelation with both actions
Figure 9 shows a typical signal for each food in exper-iment 2 In this experiment we calculate the recognitionaccuracy of food type classification by using Random Forestclassier ofWeka Tables 5 and 6 show the classification resultsThe average recognition accuracy was 896 If we did nothave a SenStick it would be hard to develop the sensorizedchopsticks that can be used in the actual environment
42 Daily Activity Monitoring by a Sensorized Belt Recentlycorpulent people are rapidly increasing and about 700 mil-lion people belong to the obesity group [18] Nowadays manyoffice workers use computers heavily and their sitting time
10 Journal of Sensors
Step 1 crop action Step 2 intake action
Lower
Upper
2 4 6 8 10 12 140Elapsed time (sec)
2 4 6 8 10 12 140Elapsed time (sec)
2 4 6 8 10 12 140Elapsed time (sec)
2 4 6 8 10 12 140Elapsed time (sec)
X
Y
Z
Synthetic
X
Y
Z
Synthetic
X
Y
Z
Synthetic
X
Y
Z
Synthetic
minus250
minus150
minus50
50
150
250
Ang
ular
velo
city
(rad
s)
minus20
minus10
01020
Acce
lera
tion
(ms
2)
minus250
minus150
minus50
50
150
250
Ang
ular
velo
city
(rad
s)
minus20
minus10
01020
Acce
lera
tion
(ms
2)
Open1 Close2 Raise3 Eat4 Drop5
1 2 3 4 5
Figure 8 Time series of sensing data in eating activity
in office increases Keeping incorrect postures during longworking time increases the risk of obesity of office workersTherefore it is necessary tomonitor the daily activity of officeworkers for correcting the bad habits
Motivated by this we have developed a sensorized beltcalled Waiston Belt [8] which aims to monitor a userrsquosdaily activities for maintaining health As products of belt-type wearable devices Welt [19] and Belty [20] are on saleThese provide the function of waist measurement activity
tracking and step counting However API for obtainingsensing data is not provided and it is not suitable for researchpurposes Generally it is hard for ordinary researchers toembed sensors battery BLE and vibrator into a small buckle-size case that can be worn in everyday life Nam et al [21]have proposed a belt-attached device for monitoring postureHowever the device has not reached a sufficient level to beused on a daily basis because of exposed wiring and uncoolappearance However by extending SenStick we succeeded
Journal of Sensors 11
C1
C2 C3
C4
Typical signal of each food
C1 rice amp dish C2 miso soup C3 noodles C4 shrimp
400
200
0
minus200
minus400
Figure 9 Eating activity recognition by sensorized chopsticks
SenStick
Waiston Belt
Figure 10 Waiston Belt based on SenStick
Table 5 Experimental result (food type)
Food type Number ofpieces of data Precision Recall 119865-Measure
C1 rice and dish 3011 0894 0945 0919C2 miso soup 2187 0907 0841 0873C3 noodles 2761 088 0896 0888C4 shrimp 2091 0907 088 0893
0896 0896 0895
to develop a stylish prototype of Waiston Belt shown inFigure 10 quickly Since the battery life of SenStick is 15 hourswith a small battery (60mAh) we can sense continuously inour daily lives As shown in Figure 11 it is possible to estimatevarious usersrsquo contexts (How much did you walk Are youusing the stairs Is posture good at work) by analyzing timeseries of sensing data obtained from the sensorized belt
An important point in this use case is that usersrsquo contextscan be estimated by combining multiple sensing data For
Table 6 Confusion matrix of food type recognition
PredictedC1 C2 C3 C4
ActualC1 rice and dish 094 004 001 000C2 miso soup 012 084 004 000C3 noodles 002 002 090 006C4 shrimp 001 001 011 088
example climbing stairs is estimated from the changes ofacceleration and atmospheric pressure The change of thehumidity data will be helpful to detect that a user enteredan air-conditioned room A SenStick makes it possible toeasily carry out research on behavior recognition with acombination of such multiple sensors
We introduce basic experiments about the posture recog-nition during deskwork by using this beltThe purpose of thisexperiment is to identify the posture type from the attitude
12 Journal of Sensors
Enter the lab
HumidityTemp
Humidity dropping
Pressure dropping
Working
Walking WalkingClimbing stairs Sitting (good posture) Sitting (bad posture)
AccsXAccsYAccsZ
60
40
20
1003
1002
1001
1000
999
25
2
15
1
05
0
minus15
minus1
minus05
Air pressure
Figure 11 Time series of sensing data in daily activity
C1 good posture C2 bad posture 1
C3 bad posture 2 C4 bad posture 3
Figure 12 Time series of sensing data in posture monitoring
angle of the belt calculated from 3 axial acceleration sensorsWe set the sensing rate as 10Hz and collected 6000 samples(10min) for each posture We assumed 4 classes (C1 goodposture C2 bad posture 1 C3 bad posture 2 and C4 badposture 3) as shown in Figure 12
In this experiment we calculated the recognition accu-racy of posture type classification by using Random Forestclassifier of Weka Tables 7 and 8 show the classificationresults The average recognition accuracy was 997 Inthe future based on the result of posture estimation we
Journal of Sensors 13
Sensorized glasses
Figure 13 Sensorized glasses
Table 7 Experimental result (posture type)
Posture type Number ofpieces of data Precision Recall 119865-Measure
C1 good posture 6000 1000 1000 1000C2 bad posture 1 6000 0994 0994 0994C3 bad posture 2 6000 0994 0994 0994C4 bad posture 3 6000 0999 0999 0999
0997 0997 0997
will develop a software application that suggests correctinghisher bad posture to a right posture during desk work
43 Indoor Localization by Sensorized Glasses We have pro-posed an indoor localization system using illuminance-basedtrilateration method [22] In this system we assume that theperson wears glasses having an illuminance sensor as shownin Figure 13 and controllable light sources are deployed in thetarget environment Our system controls lights andmeasuresthe variation of illuminance at the glasses As shown inFigure 14 the distances (1198891ndash1198893) from each light (1198971ndash1198973) to theuser (119906) are estimated from the distance-illumination modelconstructed in advance and userrsquos position is estimated basedon trilateration method using estimated distances
As a result of an experiment where three controllablelight sources were deployed in a room we confirmed thatour system could estimate the position within 05m erroron average By using SenStick we could easily develop thesensorized glasses equipped with illuminance sensor
44 Sensorized Airsoft Gun We have developed the sen-sorized airsoft gun for activity recognition in airsoft gameas shown in Figure 15 Airsoft game is a sport in whichparticipants eliminate opponents by hitting each other usingreplica weapons called airsoft guns The graph of a typicalsignal from the sensorized airsoft gun is shown in Figure 16A user can count the number of shots of a ball by attachingSenStick to his airsoft gun
45 Another Example Studentrsquos Unique Achievements andSurvey Results SenStick is also used in a lecture of IoT
Table 8 Confusion matrix of posture type recognition
PredictedC1 C2 C3 C4
ActualC1 good posture 1000 0000 0000 0000C2 bad posture 1 0000 0994 0006 0001C3 bad posture 2 0000 0006 0994 0000C4 bad posture 3 0000 0001 0000 0999
prototyping Students realize their unique IoT devices easilyby using SenSticks Here we introduce some of the developedIoT devices and survey results about the usefulness ofSenStick answered by students
Figure 17 shows a sensorized pencil with two SenSticksThe sensorized pencil is developed to recognize the charac-ters written by this pencil Even though two SenSticks areembedded to detect precise movement of pencil the size ofthe pencil is reasonable
Figure 18 shows a sensorized name tag anddigital signageSince we changed the battery from 60mAh to 900mAh thetag keeps running more than one week It aims to recordemployeesrsquo or studentsrsquo daily activities to support their healthThe digital signage informs the user who wears a sensorizedname tag about the userrsquos activity data when the user standsin front of it
Figure 19 shows a sensorized spoon used in the digitalfish catching game for children In this game the sensorizedspoon detects the scooping movement of the hand Thuschildren can catch the digital fish displayed by the projectionmapping by using sensorized spoonThis systemwas actuallyused at the universityrsquos open campus event and around100 children enjoyed this game and showed interest in IoTtechnology
The survey results obtained from the ten graduate stu-dents (ninemales and one female) who developed prototypesas mentioned above are shown in Table 9 A questionnaireis a five-grade form (5 strongly agree 4 agree a little 3neither agree nor disagree 2 disagree a little and 1 stronglydisagree) From the results of questions 1 and 2 manyparticipants think that SenStick platform is useful for IoT
14 Journal of Sensors
Position
Distance illumination model
Using trilateration method
a
Light 3
Light 2
Light 1
L = minus421d + 1496
L = minus063d + 501
05 1 15 2 25 3 35 4 450Distance between the illuminance sensor and the lighting device (m)
02468
1012141618
Illum
inan
ce v
alue
(lux
)
Distanceu
d1
d2
d3
l1 p1 x1
l3 p3 x3
l2 p2 x2
Figure 14 Indoor localization by sensorized glasses
Sensorized airsoft gun
Figure 15 Sensorized airsoft gun
research and IoT education Also the results of question 3show that many students are interested in the SenStick andwant to use it again From the results of questions 4 and 5we can see that usability and specification of SenStick arerelatively high There were opinions about the usefulness ofSenStick platform such as ldquoI could quickly try out ideas of IoTprototypingrdquo ldquoIt was easy to collect sensing datardquo and ldquoSinceSenStick is small it was easily attached tomost of the thingsrdquoOn the other hand there were opinions about improvements
such as ldquoI want to use higher sensing frequency (50ndash100Hz)rdquoldquoI want to process a sensing data on SenStickrdquo and ldquoI want tomodify the firmware itself depending on the application ofthe IoT devicerdquo
5 Conclusion
In this paper we have introduced a smart and comprehensivesensing platform called SenStick which is composed of
Journal of Sensors 15
Typical signal of airso gun
Hold a gun Ready
Shoot
1 2 3 4 5 6 70
00
30
60
91
21
51
82
12
42
73
03
33
63
94
24
54
85
15
45
76
06
36
66
97
27
57
88
18
48
79
09
39
69
910
210
510
811
111
411
712
012
312
612
913
2
minus600
minus400
minus200
0
200
400
600
800
acc_xacc_yacc_z
gyro_xgyro_y
acc_synthesis gyro_zgyro_synthesis
Figure 16 Time series of sensing data in shooting an airsoft gun
Sensorized pencil
Figure 17 Sensorized pencil
Sensorized name tagVisualization of sensor data
Figure 18 Sensorized name tag
Sensorized spoon Digital fishing game
Figure 19 Sensorized spoon
Table 9 Questionnaire results about SenStick
Question contents ScoreQ1 Do you think that SenStick will be useful for IoTresearch 49
Q2 Do you think that SenStick will be useful for IoTeducation 47
Q3 Do you want to use SenStick again 49Q4 Do you think that SenStick was easy to use 41Q5 Are you satisfied with specification of SenStick 41
hardware software and 3D case data The main purpose ofour platform is to simplify the sensorization of our personalbelongings
The hardware of SenStick is maybe one of the worldrsquossmallest sensor boards which includes 8 kinds of sensorsand BLE In addition to the hardware we also developedsupport software for managing monitoring recording andviewing the hardware By providing 3D case data all theresearchers can measure every tiny activity in daily life easilyand efficiently For an educational use it is also useful forunderstanding sensors and prototyping the desired systemincluding tiny censored things
Conflicts of Interest
The authors declare that there are no conflicts of interestregarding the publication of this paper
Acknowledgments
This work is supported in part by JST PRESTO and JSPSKAKENHI (Grant no 17J10021) and ldquoResearch and Devel-opment of Innovative Network Technologies to Create theFuturerdquo the Commissioned Research of National Instituteof Information and Communications Technology (NICT)Japan
16 Journal of Sensors
References
[1] O Brdiczka M Langet J Maisonnasse and J L CrowleyldquoDetecting human behavior models from multimodal observa-tion in a smart homerdquo IEEE Transactions on Automation Scienceand Engineering vol 6 no 4 pp 588ndash597 2009
[2] J Chen A Kam J Zhang N Liu and L Shue ldquoBathroomactivity monitoring based on soundrdquo in Pervasive ComputingThird International Conference PERVASIVE 2005 MunichGermanyMay 8-13 2005 Proceedings vol 3468 ofLectureNotesin Computer Science pp 47ndash61 Springer Berlin Germany2002
[3] G Bauer K Stockinger and P Lukowicz ldquoRecognizing theUse-Mode of Kitchen Appliances from Their Current Consump-tionrdquo in Smart Sensing and Context vol 5741 of Lecture Notesin Computer Science pp 163ndash176 Springer Berlin HeidelbergBerlin Heidelberg 2009
[4] K Ouchi and M Doi ldquoIndoor-outdoor activity recognition bya smartphonerdquo in Proceedings of the the 2012 ACM Conferencep 537 Pittsburgh Pennsylvania September 2012
[5] A Anjum and M U Ilyas ldquoActivity recognition using smart-phone sensorsrdquo in Proceedings of the IEEE 10th ConsumerCommunications and Networking Conference (CCNC rsquo13) pp914ndash919 IEEE January 2013
[6] O D Lara and M A Labrador ldquoA survey on human activityrecognition using wearable sensorsrdquo IEEE CommunicationsSurveys amp Tutorials vol 15 no 3 pp 1192ndash1209 2013
[7] ldquoJins memerdquo httpsjins-memecom[8] Y Matsuda T Ishioka T Hasegawa et al ldquoWaistonBelt A belt
for monitoring your real abdominal circumference foreverrdquo inProceedings of the ACM International Joint Conference on Perva-sive and Ubiquitous Computing and the 2015 ACM InternationalSymposium on Wearable Computers UbiComp and ISWC 2015pp 329ndash332 September 2015
[9] M Kurahashi K Murao T Terada and M Tsukamoto ldquoPer-sonal identification system based on rotation of toilet paperrollsrdquo in Proceedings of the IEEE International Conference onPervasive Computing and CommunicationsWorkshops (PerComWorkshops rsquo17) pp 521ndash526 IEEE Kona Hawaii USA March2017
[10] ldquoHapiforkrdquo httpswwwhapicom[11] Y Arakawa ldquoSenStick Sensorize every thingsrdquo in Proceedings of
the ACM International Joint Conference on Pervasive and Ubiq-uitous Computing and the 2015 ACM International SymposiumonWearable Computers UbiComp and ISWC 2015 pp 349ndash352September 2015
[12] Y Nakamura T Kanehira Y Arakawa and K YasumotoldquoSenStick 2 Ultra tiny all-in-one sensor with wireless chargingrdquoinProceedings of the 2016ACM International Joint Conference onPervasive and Ubiquitous Computing UbiComp 2016 pp 337ndash340 September 2016
[13] ldquoSensortag cc2541 (texas instruments)rdquo httpwwwticomtoolcc2541dk-sensor
[14] ldquoIot smart modulerdquo httpwwwalpscomjiotsmart[15] ldquoTsnd121 (atr-promotions)rdquo httpwwwatr-pcomproducts
TSND121html[16] ldquoWax9 (axivity)rdquo httpaxivitycomproductwax9[17] Uk biobank ldquoUk biobankrdquo httpwwwukbiobankacuk[18] J C Seidell and J Halberstadt ldquoThe global burden of obesity
and the challenges of preventionrdquo Annals of Nutrition andMetabolism vol 66 supplement 2 pp 7ndash12 2015
[19] ldquoWeltThe smart belt for fashion and healthrdquo httpswwwwelt-corpcom
[20] ldquoBelty power The belt your really deserverdquo httpswwwbeltyco
[21] H Nam J-H Kim and J-I Kim ldquoSmart Belt A wearabledevice for managing abdominal obesityrdquo in Proceedings of theInternational Conference on Big Data and Smart ComputingBigComp 2016 pp 430ndash434 January 2016
[22] K Moriya M Fujimoto Y Arakawa and K Yasumoto ldquoIndoorlocalization based on distance-illuminance model and activecontrol of lighting devicesrdquo in Proceedings of the 2016 Interna-tional Conference on Indoor Positioning and Indoor NavigationIPIN 2016 pp 1ndash6 October 2016
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal of
Volume 201
Submit your manuscripts athttpswwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
2 Journal of Sensors
Hardware
SenStick
Soware
3D data
Sensorizeevery thing
+
IoTeducation
+ Create a community3D case data
+ Share 3D data
Multiplatform(iOSAndroidPC)+ Functions for research+ Functions for education
SenStick 2+ Shorter+ Separable+ Wireless sync+ Long life
Wireless charging model+ inner+ Wireless charging+ Sealed case
Research on IoT IoT education
SenStick 1+ Ultra tiny (3 g)+ Multisensors+ BLE+ Memory+ Battery 2017
SenStick 3+ Commercialized+ nRF52 (Cortex-M4F)
Open Source
2015
2016
Figure 1 SenStick platform including hardware software and community
thingsmentioned as followsThe size of the board is relativelylarge In addition the final deliverables usually become largerbecause the baseboard requires various additional functionssuch as a charging circuit clock and storage and those areconnected via the relatively large connector As a result itis actually difficult to develop ldquosmallrdquo sensorized things thatcan be used in the actual environment Another problem issoftware for controlling the deliverable from a smartphone Itis not easy to develop a sophisticated mobile application foreach deliverable respectively Since the essence of IoT is dataanalysis sensorization of everything should be simplified andsolution for collecting data smoothly is required
Based on above-mentioned background we starteddeveloping a platform called SenStick [11] for sensorizingvarious things more easily in 2014 Our proposed SenStickis a comprehensive platform including the hardware soft-ware 3D case data and community as shown in Figure 1SenStick aims to allow researchers to sensorize everythingand also educators to teach IoT-related techniques such as
data analysis The size of the first SenStick board releasedin 2015 (SenStick 1) is 75mm (119882) times 10mm (119867) times 5mm(119863) and its weight is around 3 g including a battery On thistiny board 8 sensors (acceleration gyro magnetic light UVtemperature humidity and pressure) flash memory BLEbattery and charging circuit are densely embedded
The second version of the hardware (called SenStick 2)released in 2016 [12] became shorter 50mm (119882) times 10mm(119867) times 5mm (119863) by dividing the board into two boards andconnecting them vertically By updating the firmware thebattery life of SenStick 2 reached 15 hours by 60mAh batteryAlso wireless data synchronization function and DeviceFirmware Update (DFU) over BLE have been implementedSenStick 2 has a derivative version (wireless chargingmodel)This version has a different battery type and charging circuitfor enabling wireless charging
In 2017 the third version of the hardware (called SenStick3) was sold from a company (httpsenstickcom) Almostall the configurations of SenStick 3 are the same as SenStick
Journal of Sensors 3
2 but the board is unified into one board again like SenStick1 The main change is the upgrade of BLE chip from nRF51 tonRF52 Since the processing power is increased we can addsome calculation process on the board
In the software part support application is developedfor mobile devices (iOS and Android) and desktop PC(LinuxmacOS andWindows) Each application canmonitorand record sensing data synchronously and can configurethe parameters of each sensor The Android version cansimultaneously record a ground truth video and sensing dataThe PC version allows a user to connect multiple SenStickssimultaneously and also provides a function to use Node-RED which provides a simple programming interface
In 3D case part we designed the data of basic 3D case ofSenStick It is possible to easily design the 3D case for variousthings by using our base 3D case data
The remaining part of the paper is organized as followsWe first introduce various existing products and comparethem with SenStick in Section 2 Next we explain the detailsof our proposed platform SenStick in Section 3 In Section 4we introduce some case studies utilizing SenStick and somesurvey results answered by students who actually used Sen-Stick platform Finally we conclude the paper in Section 5
2 Related Work
We summarize competitive sensors in the market in Table 1SensorTag [13] of Texas Instruments is the most competi-
tive sensor board similar to SenStick The sensors embeddedon SensorTag are almost the same as SenStick except for UVsensor However SensorTag does not have stand-alone datarecording function and a flash memory because this sensoris designed on the premise that sensing data are always storedon the smartphone connected by BLE Since the main targetof SensorTag is environmental sensing the sensing frequencyof motion sensors is limited up to 1Hz and a rechargeablebattery is adopted Therefore it is not suitable for activitysensing of IoT research In addition the board shape size andweight are completely different SenStick is much smaller andslimmer
IoT Smart Module [14] is a small sensor board equippedwith several sensors like SensorTag Support applicationsprovide functions to cooperate with IBM Bluemix so itis possible to smoothly perform tasks from sensing ofdata to cloud analysis However support applications onlywork on Android Similar to SensorTag of Ti it adoptsnonrechargeable coin battery Although it can set highersensing frequency a user must change the coin battery atshort intervals Also logically it cannot transmit all thesensing data in case of higher sensing frequency becausethe transmission capacity of BLE is low Therefore it is notsuitable for research that requires precise data
TSND121 [15] of ATR-Promotions is another competitivesensor because it was used in some academic research Ithas 9-axis motion sensors and ambient pressure sensor andadopts Classic Bluetooth 20 for transmitting data Becauseit only focuses on monitoring activities of a human and arobot the size of the sensor is relatively large but at most7 sensors can be simultaneously connected to a control PC
Since it always requires a PC for recording sensed data it isnot suitable for sensing our daily life out of the lab
WAX9 [16] of Axivity used in UK Biobank [17] was astreaming inertialmeasurement unit (IMU)which has a largeflash memory for logging The sensor combines a MEMSaccelerometer gyro and magnetometer with a Bluetoothlow energy (BLE) compatible radio Additionally WAX9 alsofeatures a barometric pressure sensor and temperature sensorHowever it does not have some environmental sensors suchas humidity light and UV sensor and always only distributesthe Windows-compatible software for data collection andvisualization It has already been discontinued
Based on the features of other competing sensors wesummarize the requirements for SenStick platform
(1) Hardware Requirements The SenStick platform aims tosense tiny activities that cannot be sensed by existing smart-phones and wearable devices Therefore essential functionsmust be embedded in a small sensor board and each sensormust collect data in high sampling rate Since BLE commu-nication dose not guarantee reliable data communicationit is necessary to be able to record in standalone withoutrecording devices such as a smartphone In addition it isrequired that the SenStick must have rechargeable batterysystem because a coin battery is not suitable for frequent use
(2) Software Requirements The SenStick platform aims tosimplify the data collection and analysis from sensorizedthings Therefore support software should be multiplatformIt should be compatible with the various operating systemsSupport software requires not only basic function such as datarecording but also additional functions such as simultaneousrecording of sensing data frommultiple SenSticks and simpleprogramming interface for data processing
(3) 3D Case Requirements The SenStick platform aims topromote the sensorization of everything For this purposeit is desirable to provide a 3D case of SenStick that can beattached to various belongings Further 3D cases createdby someone should be shared with other users through ourwebsite
3 Comprehensive Sensing Platform forIoT Research SenStick
Our proposed platform called SenStick consists of 3 com-ponents hardware software and 3D case In the followingsubsections we explain the details of these components Thisplatform allows a user to sensorize everything easily andefficiently We believe that it supports the rapid prototypingby researchers and is useful for IoT education (a videoillustrating SenStickrsquos hardware software and 3D case datais provided in Supplementary Material available online athttpsdoiorg10115520176308302)
31 Hardware Aswe explained in Section 1 we have updatedthe hardware three times from 2015 The explanation in thefollowing part is based on SenStick 2 developed in 2016The size of the sensor board is 55mm (119882) times 10mm (119867)
4 Journal of Sensors
Table1Com
paris
onof
competitives
ensors
SenS
tick
3Se
nStic
k 2
Sens
orTa
g (C
C265
0)Io
T Sm
art M
odul
eTS
ND
121
WA
X9
Size
(mm)
50(119882
)times10
(119867)times
10(119863
)50
(119882)times
10(119867
)times10
(119863)
42(119882
)times32
(119867)times
10(119863
)44
(119882)times
27(119867
)times11(119863
)37
(119882)times
46(119867
)times12
(119863)23
(119882)times
325(119867
)times76
(119863)
Weight(g)
35
28mdash
227
Stand-alon
erecording
Itimes
timestimes
ISm
artpho
neapp
II
Itimes
timesDesktop
app
I998779
timesI
IBa
ttery
life
15(hou
r)1(year)
1(year)
6(hou
r)6(hou
r)Ba
ttery
type
Rechargeablerem
ovable
Rechargeable
Coinbatte
ryCoinbatte
ryRe
chargeable
Rechargeable
Network
BLE
BLE
BLE
Bluetooth20
BLE
Datap
rocessing
Itimes
timestimes
timestimes
I2CEx
tension
Itimes
timestimes
timestimes
Sensors
Acceleratio
nI
II
II
Gyro
II
timesI
IMagnetic
II
II
IAirpressure
II
II
ITemperature
II
Itimes
IHum
idity
II
Itimes
timesLight
II
Itimes
timesUV
Itimes
Itimes
timesMi c
timesI
timestimes
times
Journal of Sensors 5
Flash memory(back side)
50GG
60GB
Temperature amphumidity(SHT-20)
UV(VEML6070)
Pressure(LPS25HB)
Light(BH1780GLI)
Charger(BQ24090)
Micro USB
Li-Po battery
USBUART(CP2104)
9-axis(MPU-9250)
BLE(nRF51822)
Figure 2 Hardware of SenStick 2 (8 sensors BLE flash memory etc)
Table 2 Sensors embedded in SenStick 23
Type of sensor Model number Power consumption9-Axis MPU-9250 280 120583Asim37mAPressure LPS25HBTR 4 120583Asim25 120583ATemperaturehumidity SHT20 270 120583Asim330 120583AIlluminance BH1780GLI-E2 120120583Asim200 120583AUltraviolet (UV) VEML6070 100 120583Asim250 120583A
times 5mm (119863) and its weight is around 3 (g) including thebattery On this tiny board eight sensors (acceleration gyromagnetic light UV temperature humidity and pressure)flash memory (32Mbytes) BLE battery and charging circuitare densely embedded as shown in Figure 2 The modelnumber and power consumption of the sensors embedded inSenStick are listed in Table 2 All the sensors are popular chipsused worldwide and their performances are ensured by eachmanufacturer
The novelty of our hardware is the design of the sensorboard composed of the elaborated combination to satisfythe requirements presented in Section 2 This contributionbreaks down into the following aspects
(1) Combination of BLE and Flash Memory Almost allthe BLE-based devices equipped no large flash memoryto record sensing data because these devices are designedon the premise that sensing data are always recorded onsmartphones connected via BLE However the transmissionspeed of BLE is not enough for sending all the sensing dataof multiple sensors with the high sampling frequency As aresult some data will be dropped in the transmission Sincethe design of existing BLE-based devices is not appropriatefor research purpose we designed a new BLE-based sensingboard with a large flash memory which allows users torecord all the sensing data precisely Also it means that it canwork without holding a smartphone This ability is useful forsensing themotion of animals that cannot have a smartphone
(2) Selection of Powerful and Energy-Saving SoC Weselected Nordicrsquos nRF52832 SoC as SenStickrsquos microcon-troller Because the nRF52832 SoC is built around a 32-bit ARM Cortex-M4F CPU with 512 kB + 64 kB RAM itis extremely powerful Moreover the embedded 24GHztransceiver supports Bluetooth low energy and has extremelylow power consumption and it is the worldrsquos smallest classmodule Each sensor is connected via two I2C buses inde-pendently processing high-speed sensor (acceleration gyroandmagnetic) and low-speed sensor (light UV temperaturehumidity and pressure) taking into account the differencein bus occupancy time This makes it possible to realize highsampling with the motion sensors
(3) Application-Friendly Battery Type and Shape Since Sen-Stick is designed to bemounted on objects such as chopstickstoothbrushes and glasses it is desirable that the shape is thinand compact However sensors such as SensorTag equippedwith readily available coin batteries are inappropriate inshape because the footprint is square Also in the researchnonrechargeable coin batteries are costly and troublesomeFor those reasons we selected a thin and compact lithiumpolymer battery on SenStick This type of battery was harderto obtain than expected Therefore the procurement ofbattery has become the bottleneck of development
SenStickrsquos firmware is newly designed and developed byconsidering aforementioned requirements Generic Attribute(GATT) profile of SenStick provided by firmware is shownin Figure 3 The GATT server in SenStick provides threeservices (SenStick Control Service Metadata Read Serviceand Sensor Service) The SenStick Control Service providesoperation instructions such as starting and stopping sensingand logging The metadata reading service provides thefunction of reading metainformation of log data The SensorService provides functions to specify the sensing and loggingoperation of each sensor (0 acceleration 1 gyro 2 magnetic3 light 4 UV 5 temperaturehumidity 6 pressure) Also itprovides real-time sensing data reading function and log datareading function
6 Journal of Sensors
(32-)
Peripheral
GATT server
SenStick Control Service
Metadata Read Service
Read-write
Data notication
Flashmemory
(0 acceleration 1 gyro 2 magnetic 3 light
4 UV 5 temperaturehumidity 6 pressure)
GATT client
Central
~Sensor Service (0 6)
Figure 3 Generic Attribute (GATT) profile of SenStick
Table 3 BLE communication specification
Type of content SpecAdvertising packet Complete UUID flagScan response Local name (SenStick)Advertising packet interval 100ms (30 s)rarr1000ms (150 s)rarrsleepConnection interval max 80msmin 20msSupervision timeout 4000msSlave latency 0
Basically all the sensing data will be notified to a smart-phone in real time up to a transmission capacity of BLE Ifthe amount of sensing data exceeds the transmission capacitya part of sensing data will be dropped Therefore the real-time sensing data reading function is used formonitoring thestatus or for an interactive application that does not requireprecise data For precise data acquisition the log data readingfunction which guarantees the data reading order is usedThe communication specifications of BLE are described inTable 3
As shown in Table 4 the fixed capacities of a flashmemory are allocated to each sensorThe number of samplesindicates the total number of samples that can be recorded bythe allocated capacities
SenStick stops logging when the number of samples ofone sensor exceeds the storage capacity allocated for it Themaximum recording time of acceleration gyroscope andmagnetic sensor is 4 hours 40 minutes at sampling intervalof 10 milliseconds (sampling frequency 100Hz) and 14 hours10 minutes at sampling period of 30 milliseconds (samplingfrequency 33Hz) The maximum recording time of othersensors is 9 hours 26 minutes at sampling period of 200milliseconds (sampling frequency 5Hz)
32 Software We provide mobile applications for iOS andAndroid and a library for nodejs platform Common func-tions of them are operating function parameter settingfunction and real-time data view function An additional
Table 4 Allocated storage capacity per sensor
Amount of allocatedstorage capacity Samples
Acceleration 102Mbytes 17M samplesGyro 102Mbytes 17M samplesMagnetic 102Mbytes 17M samplesIlluminance 340KB 170K samplesUV (UV) 340KB 170K samplesHumidity 170KB 170K samplesAir temperature 170KB 170K samples
function of iOS application is a firmware update functionThose of Android application and nodejs applications areground truth video recording function and an ability to con-nect multiple SenSticks respectively Since all these softwareapplications are shared on GitHub as an own source softwareall the users can create their application based on their ownpurposes
Communication Protocol Here we explain the communica-tion protocol between SenStick and softwareThe explanationis based on ldquonode-SenStickrdquo which is our developed libraryfor nodejs Figure 4 shows a protocol sequence betweenSenStick and the central terminal (node-SenStick on Mac)
(I) Discovery First of all the application needs to find thesurrounding SenStick by using SenStickdiscovery() methodin our library When SenStickdiscovery() method finds aSenStick a callback function with an instance of SenStickobject containing the SenStickrsquos UUID and LocalName asarguments is called
(II) Connect If the target SenStick is found by the discoveryprocess the application is going to establish a connec-tion with it by calling SenStickprototypeconnectAndSetUp()method After establishing the connection this methodacquires the information necessary for operation from thetarget SenStick
Journal of Sensors 7
Central
Node-SenStick SenStick
SenStickdiscovery()
SenStickprototypeconnectAndSetUp()
Establish connection
SenStickprototypewriteSensorMeasurementCong()
Reect setting
SenStickprototypestartSensingAndLogging()
SenStickprototypenotifySensor()Start sensing
Select sensor to notify
Sensing data
SenStickprototypestopSensingAndLogging()
Stop sensing
Enable log notication
Select log data to notify
SenStickprototypewriteSensorLogReadoutTargetID()
Logging data
SenStickprototypedisconnect()
Disconnect
SenStickprototypenotifySensorLogData()
Periperal
(I) Discovery
(II) Connect
(III) Setting
(V) Obtaining log
(VI) Disconnect
(IV) Start sensingamp
RT receiving
Figure 4 Communication sequence between SenStick and node-SenStick
(III) Setting After establishing the connection the appli-cation sets three parameters (operation mode samplinginterval and measurement range) for each sensor by usingSenStickprototypewriteSensorMeasurementConfig()methodThe operation mode is a parameter that sets valida-tioninvalidation of sensing and logging of each sensor Thesampling interval and measurement range are parameters ofeach sensorwhichwill be changed by the sensing purpose andapplications
(IV) Start Sensing and Real-Time Data Receiving When Sen-StickprototypestartSensingAndLogging() method is calledSenStick starts sensing and logging Only valid sensorsrsquodata are transmitted via BLE and also recorded into aflash memory ID of each log is automatically assignedDuring sensing SenStick notifies sensing data in realtime To receive this notification the application callsSenStickprototypenotifySensor() method (Sensor representseach sensor name) and activates the notification for
8 Journal of Sensors
(i) Frequency(ii) Range
Device select Main view Settings Firmware update
Real-time view Log list in SenStick Download from SenStick
Selected sensorSensor name
Sensing frequency
Range of y-axis
Figure 5 iOS application for SenStick
the sensor When a notification of sensing data isreceived sensorChange event will be issued If Sen-StickprototypestopSensingAndLogging() method is calledSenStick stops sensing and logging
(V) Log Data Reading To read the log data recordedin the flash memory to the application the applicationcalls SenStickprototypenotifySensorLogData()method (Sen-sor represents each sensor name) and enables log datanotification If the application specifies a valid log ID bycalling SenStickprototypewriteSensorLogReadoutTargetID()method SenStick starts informing the sensing data recordedin the log sequentially To receive notification of log datasensorLogDataReceived event will be issued By receiving thisnotification by the application it is possible to acquire the logdata recorded in the memory
(VI) Disconnect After finishing the data acquisition Sen-Stickprototypedisconnect() method will be called to releasethe connection of SenStick And SenStick will begin advertis-ing again
User InterfaceWe explain the user interface of the applicationbased on iOS version Figure 5 shows a sophisticated mobileapplication for SenStick As shown in Figure 5 the userselects target SenStick from devices listThen the applicationshows the main view In the main view the user can selectsensors to be acquired For immediate data monitoring allthe received data via BLE can be shown in the graph view inreal time The application can control all the settings of eachsensor such as sample frequency and sensing range And alsoit has firmware update function Sensing data stored in theflash memory of SenStick is viewed from log list and it can be
downloaded to the smartphone in CSV formatThis functionprevents data loss caused by BLE communication
33 3D Case Data In order to make sensorized things anattachment case for mounting the SenStick is required Wedesigned the basic 3D case of SenStick By extending thisbasic case users can design derivative cases for chopsticksglasses and so on Actually our laboratory has succeeded indesigning cases for several things as shown in Figure 6 Allof these cases data are shared on our community site Eachcase is designed in a few minutes thanks to the basic caseFor example SenStick case for glasses is made by combiningbasic case of SenStick and open 3D data of glasses distributedby i Design Studio In addition due to the popularizationof 3D printers in the world and enhancement of tools andreferences for 3D modeling all the users can print out theirdesired SenStick cases by themselves and attach our sensor toeverything
4 Case Studies of SenStick
SenStick an ultra small sensor can be embedded into varioustiny things in our daily life Our first target is our dailybelongings such as glasses chopsticks and pens becausesensorization of daily belongings is good to intuitively under-standwhat can be sensed andwhich sensor is usefuleffectiveSenStick can upgrade normal glasses to intelligent oneswhichcan monitor a head movement posture and the amount ofsunlight received If SenStick is embedded into chopsticksand pens we can observe eating activities and handwritingactivities respectively In the future we will embed it into astick for elderlymonitoring sports gearsmusical instrument
Journal of Sensors 9
Chopsticks Tooth brush Glasses Umbrella Name card Cookware
A user can design original case data in a few minutes
Basic 3D case of SenStick
Figure 6 3D case data of SenStick
Sensorized
Chopsticks
Figure 7 Eating activity recognition by sensorized chopsticks
kitchen utensils animals stuffed toy a remote control andany kind of things that can be sensing targets In this sectionwe introduce some case studies results utilizing SenStick
41 Eating Behavior Recognition by Sensorized Chopsticks Inmany cases East Asian people eat meals with chopsticksBy analyzing the data obtained from chopsticks there is apossibility that we can estimate the eating habits includinganswers for questions like ldquohow did you eatrdquo and ldquowhatdid you eatrdquo and so forth These data are very useful formaking healthcare applications and leading a healthy lifeThis is one reason why we develop the sensorized chopsticksusing SenStick shown in Figure 7
Research goals of the eating behavior recognition by sen-sorized chopsticks are as follows G1 (behavior) recognizingthe timing and speed and detecting ldquocroprdquo and ldquointakerdquoaction G2 (food type) classifying food content like ricenoodles and so forth from the eating behavior and G3(person) identifying an individual from using chopsticks Inthis paper we show the research results on G1 and G2
We conducted basic experiments to recognize the eatingbehavior (G1 and G2) We used acceleration and gyroscopedata measured by SenStick where the sensing rate is 10HzThe purpose of experiment 1 is to understand the basic
characteristics of sensing data from sensorized chopstickwhen a participant ate chips The purpose of experiment 2 isto identify the food type from the difference in movement ofchopsticks We assumed 4 classes (C1 rice and dish C2 misosoup C3 noodles and C4 shrimp) as food types
Figure 8 shows time series of sensing data for a set ofeating actions in experiment 1 The eating actions are cropaction and intake action In crop action 6-axis motionsensors detect movement (open and close) of the chopsticksIn intake action sensors detect movement of the arm fromthe dish to the mouth Each graph in Figure 8 shows certaincorrelation with both actions
Figure 9 shows a typical signal for each food in exper-iment 2 In this experiment we calculate the recognitionaccuracy of food type classification by using Random Forestclassier ofWeka Tables 5 and 6 show the classification resultsThe average recognition accuracy was 896 If we did nothave a SenStick it would be hard to develop the sensorizedchopsticks that can be used in the actual environment
42 Daily Activity Monitoring by a Sensorized Belt Recentlycorpulent people are rapidly increasing and about 700 mil-lion people belong to the obesity group [18] Nowadays manyoffice workers use computers heavily and their sitting time
10 Journal of Sensors
Step 1 crop action Step 2 intake action
Lower
Upper
2 4 6 8 10 12 140Elapsed time (sec)
2 4 6 8 10 12 140Elapsed time (sec)
2 4 6 8 10 12 140Elapsed time (sec)
2 4 6 8 10 12 140Elapsed time (sec)
X
Y
Z
Synthetic
X
Y
Z
Synthetic
X
Y
Z
Synthetic
X
Y
Z
Synthetic
minus250
minus150
minus50
50
150
250
Ang
ular
velo
city
(rad
s)
minus20
minus10
01020
Acce
lera
tion
(ms
2)
minus250
minus150
minus50
50
150
250
Ang
ular
velo
city
(rad
s)
minus20
minus10
01020
Acce
lera
tion
(ms
2)
Open1 Close2 Raise3 Eat4 Drop5
1 2 3 4 5
Figure 8 Time series of sensing data in eating activity
in office increases Keeping incorrect postures during longworking time increases the risk of obesity of office workersTherefore it is necessary tomonitor the daily activity of officeworkers for correcting the bad habits
Motivated by this we have developed a sensorized beltcalled Waiston Belt [8] which aims to monitor a userrsquosdaily activities for maintaining health As products of belt-type wearable devices Welt [19] and Belty [20] are on saleThese provide the function of waist measurement activity
tracking and step counting However API for obtainingsensing data is not provided and it is not suitable for researchpurposes Generally it is hard for ordinary researchers toembed sensors battery BLE and vibrator into a small buckle-size case that can be worn in everyday life Nam et al [21]have proposed a belt-attached device for monitoring postureHowever the device has not reached a sufficient level to beused on a daily basis because of exposed wiring and uncoolappearance However by extending SenStick we succeeded
Journal of Sensors 11
C1
C2 C3
C4
Typical signal of each food
C1 rice amp dish C2 miso soup C3 noodles C4 shrimp
400
200
0
minus200
minus400
Figure 9 Eating activity recognition by sensorized chopsticks
SenStick
Waiston Belt
Figure 10 Waiston Belt based on SenStick
Table 5 Experimental result (food type)
Food type Number ofpieces of data Precision Recall 119865-Measure
C1 rice and dish 3011 0894 0945 0919C2 miso soup 2187 0907 0841 0873C3 noodles 2761 088 0896 0888C4 shrimp 2091 0907 088 0893
0896 0896 0895
to develop a stylish prototype of Waiston Belt shown inFigure 10 quickly Since the battery life of SenStick is 15 hourswith a small battery (60mAh) we can sense continuously inour daily lives As shown in Figure 11 it is possible to estimatevarious usersrsquo contexts (How much did you walk Are youusing the stairs Is posture good at work) by analyzing timeseries of sensing data obtained from the sensorized belt
An important point in this use case is that usersrsquo contextscan be estimated by combining multiple sensing data For
Table 6 Confusion matrix of food type recognition
PredictedC1 C2 C3 C4
ActualC1 rice and dish 094 004 001 000C2 miso soup 012 084 004 000C3 noodles 002 002 090 006C4 shrimp 001 001 011 088
example climbing stairs is estimated from the changes ofacceleration and atmospheric pressure The change of thehumidity data will be helpful to detect that a user enteredan air-conditioned room A SenStick makes it possible toeasily carry out research on behavior recognition with acombination of such multiple sensors
We introduce basic experiments about the posture recog-nition during deskwork by using this beltThe purpose of thisexperiment is to identify the posture type from the attitude
12 Journal of Sensors
Enter the lab
HumidityTemp
Humidity dropping
Pressure dropping
Working
Walking WalkingClimbing stairs Sitting (good posture) Sitting (bad posture)
AccsXAccsYAccsZ
60
40
20
1003
1002
1001
1000
999
25
2
15
1
05
0
minus15
minus1
minus05
Air pressure
Figure 11 Time series of sensing data in daily activity
C1 good posture C2 bad posture 1
C3 bad posture 2 C4 bad posture 3
Figure 12 Time series of sensing data in posture monitoring
angle of the belt calculated from 3 axial acceleration sensorsWe set the sensing rate as 10Hz and collected 6000 samples(10min) for each posture We assumed 4 classes (C1 goodposture C2 bad posture 1 C3 bad posture 2 and C4 badposture 3) as shown in Figure 12
In this experiment we calculated the recognition accu-racy of posture type classification by using Random Forestclassifier of Weka Tables 7 and 8 show the classificationresults The average recognition accuracy was 997 Inthe future based on the result of posture estimation we
Journal of Sensors 13
Sensorized glasses
Figure 13 Sensorized glasses
Table 7 Experimental result (posture type)
Posture type Number ofpieces of data Precision Recall 119865-Measure
C1 good posture 6000 1000 1000 1000C2 bad posture 1 6000 0994 0994 0994C3 bad posture 2 6000 0994 0994 0994C4 bad posture 3 6000 0999 0999 0999
0997 0997 0997
will develop a software application that suggests correctinghisher bad posture to a right posture during desk work
43 Indoor Localization by Sensorized Glasses We have pro-posed an indoor localization system using illuminance-basedtrilateration method [22] In this system we assume that theperson wears glasses having an illuminance sensor as shownin Figure 13 and controllable light sources are deployed in thetarget environment Our system controls lights andmeasuresthe variation of illuminance at the glasses As shown inFigure 14 the distances (1198891ndash1198893) from each light (1198971ndash1198973) to theuser (119906) are estimated from the distance-illumination modelconstructed in advance and userrsquos position is estimated basedon trilateration method using estimated distances
As a result of an experiment where three controllablelight sources were deployed in a room we confirmed thatour system could estimate the position within 05m erroron average By using SenStick we could easily develop thesensorized glasses equipped with illuminance sensor
44 Sensorized Airsoft Gun We have developed the sen-sorized airsoft gun for activity recognition in airsoft gameas shown in Figure 15 Airsoft game is a sport in whichparticipants eliminate opponents by hitting each other usingreplica weapons called airsoft guns The graph of a typicalsignal from the sensorized airsoft gun is shown in Figure 16A user can count the number of shots of a ball by attachingSenStick to his airsoft gun
45 Another Example Studentrsquos Unique Achievements andSurvey Results SenStick is also used in a lecture of IoT
Table 8 Confusion matrix of posture type recognition
PredictedC1 C2 C3 C4
ActualC1 good posture 1000 0000 0000 0000C2 bad posture 1 0000 0994 0006 0001C3 bad posture 2 0000 0006 0994 0000C4 bad posture 3 0000 0001 0000 0999
prototyping Students realize their unique IoT devices easilyby using SenSticks Here we introduce some of the developedIoT devices and survey results about the usefulness ofSenStick answered by students
Figure 17 shows a sensorized pencil with two SenSticksThe sensorized pencil is developed to recognize the charac-ters written by this pencil Even though two SenSticks areembedded to detect precise movement of pencil the size ofthe pencil is reasonable
Figure 18 shows a sensorized name tag anddigital signageSince we changed the battery from 60mAh to 900mAh thetag keeps running more than one week It aims to recordemployeesrsquo or studentsrsquo daily activities to support their healthThe digital signage informs the user who wears a sensorizedname tag about the userrsquos activity data when the user standsin front of it
Figure 19 shows a sensorized spoon used in the digitalfish catching game for children In this game the sensorizedspoon detects the scooping movement of the hand Thuschildren can catch the digital fish displayed by the projectionmapping by using sensorized spoonThis systemwas actuallyused at the universityrsquos open campus event and around100 children enjoyed this game and showed interest in IoTtechnology
The survey results obtained from the ten graduate stu-dents (ninemales and one female) who developed prototypesas mentioned above are shown in Table 9 A questionnaireis a five-grade form (5 strongly agree 4 agree a little 3neither agree nor disagree 2 disagree a little and 1 stronglydisagree) From the results of questions 1 and 2 manyparticipants think that SenStick platform is useful for IoT
14 Journal of Sensors
Position
Distance illumination model
Using trilateration method
a
Light 3
Light 2
Light 1
L = minus421d + 1496
L = minus063d + 501
05 1 15 2 25 3 35 4 450Distance between the illuminance sensor and the lighting device (m)
02468
1012141618
Illum
inan
ce v
alue
(lux
)
Distanceu
d1
d2
d3
l1 p1 x1
l3 p3 x3
l2 p2 x2
Figure 14 Indoor localization by sensorized glasses
Sensorized airsoft gun
Figure 15 Sensorized airsoft gun
research and IoT education Also the results of question 3show that many students are interested in the SenStick andwant to use it again From the results of questions 4 and 5we can see that usability and specification of SenStick arerelatively high There were opinions about the usefulness ofSenStick platform such as ldquoI could quickly try out ideas of IoTprototypingrdquo ldquoIt was easy to collect sensing datardquo and ldquoSinceSenStick is small it was easily attached tomost of the thingsrdquoOn the other hand there were opinions about improvements
such as ldquoI want to use higher sensing frequency (50ndash100Hz)rdquoldquoI want to process a sensing data on SenStickrdquo and ldquoI want tomodify the firmware itself depending on the application ofthe IoT devicerdquo
5 Conclusion
In this paper we have introduced a smart and comprehensivesensing platform called SenStick which is composed of
Journal of Sensors 15
Typical signal of airso gun
Hold a gun Ready
Shoot
1 2 3 4 5 6 70
00
30
60
91
21
51
82
12
42
73
03
33
63
94
24
54
85
15
45
76
06
36
66
97
27
57
88
18
48
79
09
39
69
910
210
510
811
111
411
712
012
312
612
913
2
minus600
minus400
minus200
0
200
400
600
800
acc_xacc_yacc_z
gyro_xgyro_y
acc_synthesis gyro_zgyro_synthesis
Figure 16 Time series of sensing data in shooting an airsoft gun
Sensorized pencil
Figure 17 Sensorized pencil
Sensorized name tagVisualization of sensor data
Figure 18 Sensorized name tag
Sensorized spoon Digital fishing game
Figure 19 Sensorized spoon
Table 9 Questionnaire results about SenStick
Question contents ScoreQ1 Do you think that SenStick will be useful for IoTresearch 49
Q2 Do you think that SenStick will be useful for IoTeducation 47
Q3 Do you want to use SenStick again 49Q4 Do you think that SenStick was easy to use 41Q5 Are you satisfied with specification of SenStick 41
hardware software and 3D case data The main purpose ofour platform is to simplify the sensorization of our personalbelongings
The hardware of SenStick is maybe one of the worldrsquossmallest sensor boards which includes 8 kinds of sensorsand BLE In addition to the hardware we also developedsupport software for managing monitoring recording andviewing the hardware By providing 3D case data all theresearchers can measure every tiny activity in daily life easilyand efficiently For an educational use it is also useful forunderstanding sensors and prototyping the desired systemincluding tiny censored things
Conflicts of Interest
The authors declare that there are no conflicts of interestregarding the publication of this paper
Acknowledgments
This work is supported in part by JST PRESTO and JSPSKAKENHI (Grant no 17J10021) and ldquoResearch and Devel-opment of Innovative Network Technologies to Create theFuturerdquo the Commissioned Research of National Instituteof Information and Communications Technology (NICT)Japan
16 Journal of Sensors
References
[1] O Brdiczka M Langet J Maisonnasse and J L CrowleyldquoDetecting human behavior models from multimodal observa-tion in a smart homerdquo IEEE Transactions on Automation Scienceand Engineering vol 6 no 4 pp 588ndash597 2009
[2] J Chen A Kam J Zhang N Liu and L Shue ldquoBathroomactivity monitoring based on soundrdquo in Pervasive ComputingThird International Conference PERVASIVE 2005 MunichGermanyMay 8-13 2005 Proceedings vol 3468 ofLectureNotesin Computer Science pp 47ndash61 Springer Berlin Germany2002
[3] G Bauer K Stockinger and P Lukowicz ldquoRecognizing theUse-Mode of Kitchen Appliances from Their Current Consump-tionrdquo in Smart Sensing and Context vol 5741 of Lecture Notesin Computer Science pp 163ndash176 Springer Berlin HeidelbergBerlin Heidelberg 2009
[4] K Ouchi and M Doi ldquoIndoor-outdoor activity recognition bya smartphonerdquo in Proceedings of the the 2012 ACM Conferencep 537 Pittsburgh Pennsylvania September 2012
[5] A Anjum and M U Ilyas ldquoActivity recognition using smart-phone sensorsrdquo in Proceedings of the IEEE 10th ConsumerCommunications and Networking Conference (CCNC rsquo13) pp914ndash919 IEEE January 2013
[6] O D Lara and M A Labrador ldquoA survey on human activityrecognition using wearable sensorsrdquo IEEE CommunicationsSurveys amp Tutorials vol 15 no 3 pp 1192ndash1209 2013
[7] ldquoJins memerdquo httpsjins-memecom[8] Y Matsuda T Ishioka T Hasegawa et al ldquoWaistonBelt A belt
for monitoring your real abdominal circumference foreverrdquo inProceedings of the ACM International Joint Conference on Perva-sive and Ubiquitous Computing and the 2015 ACM InternationalSymposium on Wearable Computers UbiComp and ISWC 2015pp 329ndash332 September 2015
[9] M Kurahashi K Murao T Terada and M Tsukamoto ldquoPer-sonal identification system based on rotation of toilet paperrollsrdquo in Proceedings of the IEEE International Conference onPervasive Computing and CommunicationsWorkshops (PerComWorkshops rsquo17) pp 521ndash526 IEEE Kona Hawaii USA March2017
[10] ldquoHapiforkrdquo httpswwwhapicom[11] Y Arakawa ldquoSenStick Sensorize every thingsrdquo in Proceedings of
the ACM International Joint Conference on Pervasive and Ubiq-uitous Computing and the 2015 ACM International SymposiumonWearable Computers UbiComp and ISWC 2015 pp 349ndash352September 2015
[12] Y Nakamura T Kanehira Y Arakawa and K YasumotoldquoSenStick 2 Ultra tiny all-in-one sensor with wireless chargingrdquoinProceedings of the 2016ACM International Joint Conference onPervasive and Ubiquitous Computing UbiComp 2016 pp 337ndash340 September 2016
[13] ldquoSensortag cc2541 (texas instruments)rdquo httpwwwticomtoolcc2541dk-sensor
[14] ldquoIot smart modulerdquo httpwwwalpscomjiotsmart[15] ldquoTsnd121 (atr-promotions)rdquo httpwwwatr-pcomproducts
TSND121html[16] ldquoWax9 (axivity)rdquo httpaxivitycomproductwax9[17] Uk biobank ldquoUk biobankrdquo httpwwwukbiobankacuk[18] J C Seidell and J Halberstadt ldquoThe global burden of obesity
and the challenges of preventionrdquo Annals of Nutrition andMetabolism vol 66 supplement 2 pp 7ndash12 2015
[19] ldquoWeltThe smart belt for fashion and healthrdquo httpswwwwelt-corpcom
[20] ldquoBelty power The belt your really deserverdquo httpswwwbeltyco
[21] H Nam J-H Kim and J-I Kim ldquoSmart Belt A wearabledevice for managing abdominal obesityrdquo in Proceedings of theInternational Conference on Big Data and Smart ComputingBigComp 2016 pp 430ndash434 January 2016
[22] K Moriya M Fujimoto Y Arakawa and K Yasumoto ldquoIndoorlocalization based on distance-illuminance model and activecontrol of lighting devicesrdquo in Proceedings of the 2016 Interna-tional Conference on Indoor Positioning and Indoor NavigationIPIN 2016 pp 1ndash6 October 2016
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal of
Volume 201
Submit your manuscripts athttpswwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
Journal of Sensors 3
2 but the board is unified into one board again like SenStick1 The main change is the upgrade of BLE chip from nRF51 tonRF52 Since the processing power is increased we can addsome calculation process on the board
In the software part support application is developedfor mobile devices (iOS and Android) and desktop PC(LinuxmacOS andWindows) Each application canmonitorand record sensing data synchronously and can configurethe parameters of each sensor The Android version cansimultaneously record a ground truth video and sensing dataThe PC version allows a user to connect multiple SenStickssimultaneously and also provides a function to use Node-RED which provides a simple programming interface
In 3D case part we designed the data of basic 3D case ofSenStick It is possible to easily design the 3D case for variousthings by using our base 3D case data
The remaining part of the paper is organized as followsWe first introduce various existing products and comparethem with SenStick in Section 2 Next we explain the detailsof our proposed platform SenStick in Section 3 In Section 4we introduce some case studies utilizing SenStick and somesurvey results answered by students who actually used Sen-Stick platform Finally we conclude the paper in Section 5
2 Related Work
We summarize competitive sensors in the market in Table 1SensorTag [13] of Texas Instruments is the most competi-
tive sensor board similar to SenStick The sensors embeddedon SensorTag are almost the same as SenStick except for UVsensor However SensorTag does not have stand-alone datarecording function and a flash memory because this sensoris designed on the premise that sensing data are always storedon the smartphone connected by BLE Since the main targetof SensorTag is environmental sensing the sensing frequencyof motion sensors is limited up to 1Hz and a rechargeablebattery is adopted Therefore it is not suitable for activitysensing of IoT research In addition the board shape size andweight are completely different SenStick is much smaller andslimmer
IoT Smart Module [14] is a small sensor board equippedwith several sensors like SensorTag Support applicationsprovide functions to cooperate with IBM Bluemix so itis possible to smoothly perform tasks from sensing ofdata to cloud analysis However support applications onlywork on Android Similar to SensorTag of Ti it adoptsnonrechargeable coin battery Although it can set highersensing frequency a user must change the coin battery atshort intervals Also logically it cannot transmit all thesensing data in case of higher sensing frequency becausethe transmission capacity of BLE is low Therefore it is notsuitable for research that requires precise data
TSND121 [15] of ATR-Promotions is another competitivesensor because it was used in some academic research Ithas 9-axis motion sensors and ambient pressure sensor andadopts Classic Bluetooth 20 for transmitting data Becauseit only focuses on monitoring activities of a human and arobot the size of the sensor is relatively large but at most7 sensors can be simultaneously connected to a control PC
Since it always requires a PC for recording sensed data it isnot suitable for sensing our daily life out of the lab
WAX9 [16] of Axivity used in UK Biobank [17] was astreaming inertialmeasurement unit (IMU)which has a largeflash memory for logging The sensor combines a MEMSaccelerometer gyro and magnetometer with a Bluetoothlow energy (BLE) compatible radio Additionally WAX9 alsofeatures a barometric pressure sensor and temperature sensorHowever it does not have some environmental sensors suchas humidity light and UV sensor and always only distributesthe Windows-compatible software for data collection andvisualization It has already been discontinued
Based on the features of other competing sensors wesummarize the requirements for SenStick platform
(1) Hardware Requirements The SenStick platform aims tosense tiny activities that cannot be sensed by existing smart-phones and wearable devices Therefore essential functionsmust be embedded in a small sensor board and each sensormust collect data in high sampling rate Since BLE commu-nication dose not guarantee reliable data communicationit is necessary to be able to record in standalone withoutrecording devices such as a smartphone In addition it isrequired that the SenStick must have rechargeable batterysystem because a coin battery is not suitable for frequent use
(2) Software Requirements The SenStick platform aims tosimplify the data collection and analysis from sensorizedthings Therefore support software should be multiplatformIt should be compatible with the various operating systemsSupport software requires not only basic function such as datarecording but also additional functions such as simultaneousrecording of sensing data frommultiple SenSticks and simpleprogramming interface for data processing
(3) 3D Case Requirements The SenStick platform aims topromote the sensorization of everything For this purposeit is desirable to provide a 3D case of SenStick that can beattached to various belongings Further 3D cases createdby someone should be shared with other users through ourwebsite
3 Comprehensive Sensing Platform forIoT Research SenStick
Our proposed platform called SenStick consists of 3 com-ponents hardware software and 3D case In the followingsubsections we explain the details of these components Thisplatform allows a user to sensorize everything easily andefficiently We believe that it supports the rapid prototypingby researchers and is useful for IoT education (a videoillustrating SenStickrsquos hardware software and 3D case datais provided in Supplementary Material available online athttpsdoiorg10115520176308302)
31 Hardware Aswe explained in Section 1 we have updatedthe hardware three times from 2015 The explanation in thefollowing part is based on SenStick 2 developed in 2016The size of the sensor board is 55mm (119882) times 10mm (119867)
4 Journal of Sensors
Table1Com
paris
onof
competitives
ensors
SenS
tick
3Se
nStic
k 2
Sens
orTa
g (C
C265
0)Io
T Sm
art M
odul
eTS
ND
121
WA
X9
Size
(mm)
50(119882
)times10
(119867)times
10(119863
)50
(119882)times
10(119867
)times10
(119863)
42(119882
)times32
(119867)times
10(119863
)44
(119882)times
27(119867
)times11(119863
)37
(119882)times
46(119867
)times12
(119863)23
(119882)times
325(119867
)times76
(119863)
Weight(g)
35
28mdash
227
Stand-alon
erecording
Itimes
timestimes
ISm
artpho
neapp
II
Itimes
timesDesktop
app
I998779
timesI
IBa
ttery
life
15(hou
r)1(year)
1(year)
6(hou
r)6(hou
r)Ba
ttery
type
Rechargeablerem
ovable
Rechargeable
Coinbatte
ryCoinbatte
ryRe
chargeable
Rechargeable
Network
BLE
BLE
BLE
Bluetooth20
BLE
Datap
rocessing
Itimes
timestimes
timestimes
I2CEx
tension
Itimes
timestimes
timestimes
Sensors
Acceleratio
nI
II
II
Gyro
II
timesI
IMagnetic
II
II
IAirpressure
II
II
ITemperature
II
Itimes
IHum
idity
II
Itimes
timesLight
II
Itimes
timesUV
Itimes
Itimes
timesMi c
timesI
timestimes
times
Journal of Sensors 5
Flash memory(back side)
50GG
60GB
Temperature amphumidity(SHT-20)
UV(VEML6070)
Pressure(LPS25HB)
Light(BH1780GLI)
Charger(BQ24090)
Micro USB
Li-Po battery
USBUART(CP2104)
9-axis(MPU-9250)
BLE(nRF51822)
Figure 2 Hardware of SenStick 2 (8 sensors BLE flash memory etc)
Table 2 Sensors embedded in SenStick 23
Type of sensor Model number Power consumption9-Axis MPU-9250 280 120583Asim37mAPressure LPS25HBTR 4 120583Asim25 120583ATemperaturehumidity SHT20 270 120583Asim330 120583AIlluminance BH1780GLI-E2 120120583Asim200 120583AUltraviolet (UV) VEML6070 100 120583Asim250 120583A
times 5mm (119863) and its weight is around 3 (g) including thebattery On this tiny board eight sensors (acceleration gyromagnetic light UV temperature humidity and pressure)flash memory (32Mbytes) BLE battery and charging circuitare densely embedded as shown in Figure 2 The modelnumber and power consumption of the sensors embedded inSenStick are listed in Table 2 All the sensors are popular chipsused worldwide and their performances are ensured by eachmanufacturer
The novelty of our hardware is the design of the sensorboard composed of the elaborated combination to satisfythe requirements presented in Section 2 This contributionbreaks down into the following aspects
(1) Combination of BLE and Flash Memory Almost allthe BLE-based devices equipped no large flash memoryto record sensing data because these devices are designedon the premise that sensing data are always recorded onsmartphones connected via BLE However the transmissionspeed of BLE is not enough for sending all the sensing dataof multiple sensors with the high sampling frequency As aresult some data will be dropped in the transmission Sincethe design of existing BLE-based devices is not appropriatefor research purpose we designed a new BLE-based sensingboard with a large flash memory which allows users torecord all the sensing data precisely Also it means that it canwork without holding a smartphone This ability is useful forsensing themotion of animals that cannot have a smartphone
(2) Selection of Powerful and Energy-Saving SoC Weselected Nordicrsquos nRF52832 SoC as SenStickrsquos microcon-troller Because the nRF52832 SoC is built around a 32-bit ARM Cortex-M4F CPU with 512 kB + 64 kB RAM itis extremely powerful Moreover the embedded 24GHztransceiver supports Bluetooth low energy and has extremelylow power consumption and it is the worldrsquos smallest classmodule Each sensor is connected via two I2C buses inde-pendently processing high-speed sensor (acceleration gyroandmagnetic) and low-speed sensor (light UV temperaturehumidity and pressure) taking into account the differencein bus occupancy time This makes it possible to realize highsampling with the motion sensors
(3) Application-Friendly Battery Type and Shape Since Sen-Stick is designed to bemounted on objects such as chopstickstoothbrushes and glasses it is desirable that the shape is thinand compact However sensors such as SensorTag equippedwith readily available coin batteries are inappropriate inshape because the footprint is square Also in the researchnonrechargeable coin batteries are costly and troublesomeFor those reasons we selected a thin and compact lithiumpolymer battery on SenStick This type of battery was harderto obtain than expected Therefore the procurement ofbattery has become the bottleneck of development
SenStickrsquos firmware is newly designed and developed byconsidering aforementioned requirements Generic Attribute(GATT) profile of SenStick provided by firmware is shownin Figure 3 The GATT server in SenStick provides threeservices (SenStick Control Service Metadata Read Serviceand Sensor Service) The SenStick Control Service providesoperation instructions such as starting and stopping sensingand logging The metadata reading service provides thefunction of reading metainformation of log data The SensorService provides functions to specify the sensing and loggingoperation of each sensor (0 acceleration 1 gyro 2 magnetic3 light 4 UV 5 temperaturehumidity 6 pressure) Also itprovides real-time sensing data reading function and log datareading function
6 Journal of Sensors
(32-)
Peripheral
GATT server
SenStick Control Service
Metadata Read Service
Read-write
Data notication
Flashmemory
(0 acceleration 1 gyro 2 magnetic 3 light
4 UV 5 temperaturehumidity 6 pressure)
GATT client
Central
~Sensor Service (0 6)
Figure 3 Generic Attribute (GATT) profile of SenStick
Table 3 BLE communication specification
Type of content SpecAdvertising packet Complete UUID flagScan response Local name (SenStick)Advertising packet interval 100ms (30 s)rarr1000ms (150 s)rarrsleepConnection interval max 80msmin 20msSupervision timeout 4000msSlave latency 0
Basically all the sensing data will be notified to a smart-phone in real time up to a transmission capacity of BLE Ifthe amount of sensing data exceeds the transmission capacitya part of sensing data will be dropped Therefore the real-time sensing data reading function is used formonitoring thestatus or for an interactive application that does not requireprecise data For precise data acquisition the log data readingfunction which guarantees the data reading order is usedThe communication specifications of BLE are described inTable 3
As shown in Table 4 the fixed capacities of a flashmemory are allocated to each sensorThe number of samplesindicates the total number of samples that can be recorded bythe allocated capacities
SenStick stops logging when the number of samples ofone sensor exceeds the storage capacity allocated for it Themaximum recording time of acceleration gyroscope andmagnetic sensor is 4 hours 40 minutes at sampling intervalof 10 milliseconds (sampling frequency 100Hz) and 14 hours10 minutes at sampling period of 30 milliseconds (samplingfrequency 33Hz) The maximum recording time of othersensors is 9 hours 26 minutes at sampling period of 200milliseconds (sampling frequency 5Hz)
32 Software We provide mobile applications for iOS andAndroid and a library for nodejs platform Common func-tions of them are operating function parameter settingfunction and real-time data view function An additional
Table 4 Allocated storage capacity per sensor
Amount of allocatedstorage capacity Samples
Acceleration 102Mbytes 17M samplesGyro 102Mbytes 17M samplesMagnetic 102Mbytes 17M samplesIlluminance 340KB 170K samplesUV (UV) 340KB 170K samplesHumidity 170KB 170K samplesAir temperature 170KB 170K samples
function of iOS application is a firmware update functionThose of Android application and nodejs applications areground truth video recording function and an ability to con-nect multiple SenSticks respectively Since all these softwareapplications are shared on GitHub as an own source softwareall the users can create their application based on their ownpurposes
Communication Protocol Here we explain the communica-tion protocol between SenStick and softwareThe explanationis based on ldquonode-SenStickrdquo which is our developed libraryfor nodejs Figure 4 shows a protocol sequence betweenSenStick and the central terminal (node-SenStick on Mac)
(I) Discovery First of all the application needs to find thesurrounding SenStick by using SenStickdiscovery() methodin our library When SenStickdiscovery() method finds aSenStick a callback function with an instance of SenStickobject containing the SenStickrsquos UUID and LocalName asarguments is called
(II) Connect If the target SenStick is found by the discoveryprocess the application is going to establish a connec-tion with it by calling SenStickprototypeconnectAndSetUp()method After establishing the connection this methodacquires the information necessary for operation from thetarget SenStick
Journal of Sensors 7
Central
Node-SenStick SenStick
SenStickdiscovery()
SenStickprototypeconnectAndSetUp()
Establish connection
SenStickprototypewriteSensorMeasurementCong()
Reect setting
SenStickprototypestartSensingAndLogging()
SenStickprototypenotifySensor()Start sensing
Select sensor to notify
Sensing data
SenStickprototypestopSensingAndLogging()
Stop sensing
Enable log notication
Select log data to notify
SenStickprototypewriteSensorLogReadoutTargetID()
Logging data
SenStickprototypedisconnect()
Disconnect
SenStickprototypenotifySensorLogData()
Periperal
(I) Discovery
(II) Connect
(III) Setting
(V) Obtaining log
(VI) Disconnect
(IV) Start sensingamp
RT receiving
Figure 4 Communication sequence between SenStick and node-SenStick
(III) Setting After establishing the connection the appli-cation sets three parameters (operation mode samplinginterval and measurement range) for each sensor by usingSenStickprototypewriteSensorMeasurementConfig()methodThe operation mode is a parameter that sets valida-tioninvalidation of sensing and logging of each sensor Thesampling interval and measurement range are parameters ofeach sensorwhichwill be changed by the sensing purpose andapplications
(IV) Start Sensing and Real-Time Data Receiving When Sen-StickprototypestartSensingAndLogging() method is calledSenStick starts sensing and logging Only valid sensorsrsquodata are transmitted via BLE and also recorded into aflash memory ID of each log is automatically assignedDuring sensing SenStick notifies sensing data in realtime To receive this notification the application callsSenStickprototypenotifySensor() method (Sensor representseach sensor name) and activates the notification for
8 Journal of Sensors
(i) Frequency(ii) Range
Device select Main view Settings Firmware update
Real-time view Log list in SenStick Download from SenStick
Selected sensorSensor name
Sensing frequency
Range of y-axis
Figure 5 iOS application for SenStick
the sensor When a notification of sensing data isreceived sensorChange event will be issued If Sen-StickprototypestopSensingAndLogging() method is calledSenStick stops sensing and logging
(V) Log Data Reading To read the log data recordedin the flash memory to the application the applicationcalls SenStickprototypenotifySensorLogData()method (Sen-sor represents each sensor name) and enables log datanotification If the application specifies a valid log ID bycalling SenStickprototypewriteSensorLogReadoutTargetID()method SenStick starts informing the sensing data recordedin the log sequentially To receive notification of log datasensorLogDataReceived event will be issued By receiving thisnotification by the application it is possible to acquire the logdata recorded in the memory
(VI) Disconnect After finishing the data acquisition Sen-Stickprototypedisconnect() method will be called to releasethe connection of SenStick And SenStick will begin advertis-ing again
User InterfaceWe explain the user interface of the applicationbased on iOS version Figure 5 shows a sophisticated mobileapplication for SenStick As shown in Figure 5 the userselects target SenStick from devices listThen the applicationshows the main view In the main view the user can selectsensors to be acquired For immediate data monitoring allthe received data via BLE can be shown in the graph view inreal time The application can control all the settings of eachsensor such as sample frequency and sensing range And alsoit has firmware update function Sensing data stored in theflash memory of SenStick is viewed from log list and it can be
downloaded to the smartphone in CSV formatThis functionprevents data loss caused by BLE communication
33 3D Case Data In order to make sensorized things anattachment case for mounting the SenStick is required Wedesigned the basic 3D case of SenStick By extending thisbasic case users can design derivative cases for chopsticksglasses and so on Actually our laboratory has succeeded indesigning cases for several things as shown in Figure 6 Allof these cases data are shared on our community site Eachcase is designed in a few minutes thanks to the basic caseFor example SenStick case for glasses is made by combiningbasic case of SenStick and open 3D data of glasses distributedby i Design Studio In addition due to the popularizationof 3D printers in the world and enhancement of tools andreferences for 3D modeling all the users can print out theirdesired SenStick cases by themselves and attach our sensor toeverything
4 Case Studies of SenStick
SenStick an ultra small sensor can be embedded into varioustiny things in our daily life Our first target is our dailybelongings such as glasses chopsticks and pens becausesensorization of daily belongings is good to intuitively under-standwhat can be sensed andwhich sensor is usefuleffectiveSenStick can upgrade normal glasses to intelligent oneswhichcan monitor a head movement posture and the amount ofsunlight received If SenStick is embedded into chopsticksand pens we can observe eating activities and handwritingactivities respectively In the future we will embed it into astick for elderlymonitoring sports gearsmusical instrument
Journal of Sensors 9
Chopsticks Tooth brush Glasses Umbrella Name card Cookware
A user can design original case data in a few minutes
Basic 3D case of SenStick
Figure 6 3D case data of SenStick
Sensorized
Chopsticks
Figure 7 Eating activity recognition by sensorized chopsticks
kitchen utensils animals stuffed toy a remote control andany kind of things that can be sensing targets In this sectionwe introduce some case studies results utilizing SenStick
41 Eating Behavior Recognition by Sensorized Chopsticks Inmany cases East Asian people eat meals with chopsticksBy analyzing the data obtained from chopsticks there is apossibility that we can estimate the eating habits includinganswers for questions like ldquohow did you eatrdquo and ldquowhatdid you eatrdquo and so forth These data are very useful formaking healthcare applications and leading a healthy lifeThis is one reason why we develop the sensorized chopsticksusing SenStick shown in Figure 7
Research goals of the eating behavior recognition by sen-sorized chopsticks are as follows G1 (behavior) recognizingthe timing and speed and detecting ldquocroprdquo and ldquointakerdquoaction G2 (food type) classifying food content like ricenoodles and so forth from the eating behavior and G3(person) identifying an individual from using chopsticks Inthis paper we show the research results on G1 and G2
We conducted basic experiments to recognize the eatingbehavior (G1 and G2) We used acceleration and gyroscopedata measured by SenStick where the sensing rate is 10HzThe purpose of experiment 1 is to understand the basic
characteristics of sensing data from sensorized chopstickwhen a participant ate chips The purpose of experiment 2 isto identify the food type from the difference in movement ofchopsticks We assumed 4 classes (C1 rice and dish C2 misosoup C3 noodles and C4 shrimp) as food types
Figure 8 shows time series of sensing data for a set ofeating actions in experiment 1 The eating actions are cropaction and intake action In crop action 6-axis motionsensors detect movement (open and close) of the chopsticksIn intake action sensors detect movement of the arm fromthe dish to the mouth Each graph in Figure 8 shows certaincorrelation with both actions
Figure 9 shows a typical signal for each food in exper-iment 2 In this experiment we calculate the recognitionaccuracy of food type classification by using Random Forestclassier ofWeka Tables 5 and 6 show the classification resultsThe average recognition accuracy was 896 If we did nothave a SenStick it would be hard to develop the sensorizedchopsticks that can be used in the actual environment
42 Daily Activity Monitoring by a Sensorized Belt Recentlycorpulent people are rapidly increasing and about 700 mil-lion people belong to the obesity group [18] Nowadays manyoffice workers use computers heavily and their sitting time
10 Journal of Sensors
Step 1 crop action Step 2 intake action
Lower
Upper
2 4 6 8 10 12 140Elapsed time (sec)
2 4 6 8 10 12 140Elapsed time (sec)
2 4 6 8 10 12 140Elapsed time (sec)
2 4 6 8 10 12 140Elapsed time (sec)
X
Y
Z
Synthetic
X
Y
Z
Synthetic
X
Y
Z
Synthetic
X
Y
Z
Synthetic
minus250
minus150
minus50
50
150
250
Ang
ular
velo
city
(rad
s)
minus20
minus10
01020
Acce
lera
tion
(ms
2)
minus250
minus150
minus50
50
150
250
Ang
ular
velo
city
(rad
s)
minus20
minus10
01020
Acce
lera
tion
(ms
2)
Open1 Close2 Raise3 Eat4 Drop5
1 2 3 4 5
Figure 8 Time series of sensing data in eating activity
in office increases Keeping incorrect postures during longworking time increases the risk of obesity of office workersTherefore it is necessary tomonitor the daily activity of officeworkers for correcting the bad habits
Motivated by this we have developed a sensorized beltcalled Waiston Belt [8] which aims to monitor a userrsquosdaily activities for maintaining health As products of belt-type wearable devices Welt [19] and Belty [20] are on saleThese provide the function of waist measurement activity
tracking and step counting However API for obtainingsensing data is not provided and it is not suitable for researchpurposes Generally it is hard for ordinary researchers toembed sensors battery BLE and vibrator into a small buckle-size case that can be worn in everyday life Nam et al [21]have proposed a belt-attached device for monitoring postureHowever the device has not reached a sufficient level to beused on a daily basis because of exposed wiring and uncoolappearance However by extending SenStick we succeeded
Journal of Sensors 11
C1
C2 C3
C4
Typical signal of each food
C1 rice amp dish C2 miso soup C3 noodles C4 shrimp
400
200
0
minus200
minus400
Figure 9 Eating activity recognition by sensorized chopsticks
SenStick
Waiston Belt
Figure 10 Waiston Belt based on SenStick
Table 5 Experimental result (food type)
Food type Number ofpieces of data Precision Recall 119865-Measure
C1 rice and dish 3011 0894 0945 0919C2 miso soup 2187 0907 0841 0873C3 noodles 2761 088 0896 0888C4 shrimp 2091 0907 088 0893
0896 0896 0895
to develop a stylish prototype of Waiston Belt shown inFigure 10 quickly Since the battery life of SenStick is 15 hourswith a small battery (60mAh) we can sense continuously inour daily lives As shown in Figure 11 it is possible to estimatevarious usersrsquo contexts (How much did you walk Are youusing the stairs Is posture good at work) by analyzing timeseries of sensing data obtained from the sensorized belt
An important point in this use case is that usersrsquo contextscan be estimated by combining multiple sensing data For
Table 6 Confusion matrix of food type recognition
PredictedC1 C2 C3 C4
ActualC1 rice and dish 094 004 001 000C2 miso soup 012 084 004 000C3 noodles 002 002 090 006C4 shrimp 001 001 011 088
example climbing stairs is estimated from the changes ofacceleration and atmospheric pressure The change of thehumidity data will be helpful to detect that a user enteredan air-conditioned room A SenStick makes it possible toeasily carry out research on behavior recognition with acombination of such multiple sensors
We introduce basic experiments about the posture recog-nition during deskwork by using this beltThe purpose of thisexperiment is to identify the posture type from the attitude
12 Journal of Sensors
Enter the lab
HumidityTemp
Humidity dropping
Pressure dropping
Working
Walking WalkingClimbing stairs Sitting (good posture) Sitting (bad posture)
AccsXAccsYAccsZ
60
40
20
1003
1002
1001
1000
999
25
2
15
1
05
0
minus15
minus1
minus05
Air pressure
Figure 11 Time series of sensing data in daily activity
C1 good posture C2 bad posture 1
C3 bad posture 2 C4 bad posture 3
Figure 12 Time series of sensing data in posture monitoring
angle of the belt calculated from 3 axial acceleration sensorsWe set the sensing rate as 10Hz and collected 6000 samples(10min) for each posture We assumed 4 classes (C1 goodposture C2 bad posture 1 C3 bad posture 2 and C4 badposture 3) as shown in Figure 12
In this experiment we calculated the recognition accu-racy of posture type classification by using Random Forestclassifier of Weka Tables 7 and 8 show the classificationresults The average recognition accuracy was 997 Inthe future based on the result of posture estimation we
Journal of Sensors 13
Sensorized glasses
Figure 13 Sensorized glasses
Table 7 Experimental result (posture type)
Posture type Number ofpieces of data Precision Recall 119865-Measure
C1 good posture 6000 1000 1000 1000C2 bad posture 1 6000 0994 0994 0994C3 bad posture 2 6000 0994 0994 0994C4 bad posture 3 6000 0999 0999 0999
0997 0997 0997
will develop a software application that suggests correctinghisher bad posture to a right posture during desk work
43 Indoor Localization by Sensorized Glasses We have pro-posed an indoor localization system using illuminance-basedtrilateration method [22] In this system we assume that theperson wears glasses having an illuminance sensor as shownin Figure 13 and controllable light sources are deployed in thetarget environment Our system controls lights andmeasuresthe variation of illuminance at the glasses As shown inFigure 14 the distances (1198891ndash1198893) from each light (1198971ndash1198973) to theuser (119906) are estimated from the distance-illumination modelconstructed in advance and userrsquos position is estimated basedon trilateration method using estimated distances
As a result of an experiment where three controllablelight sources were deployed in a room we confirmed thatour system could estimate the position within 05m erroron average By using SenStick we could easily develop thesensorized glasses equipped with illuminance sensor
44 Sensorized Airsoft Gun We have developed the sen-sorized airsoft gun for activity recognition in airsoft gameas shown in Figure 15 Airsoft game is a sport in whichparticipants eliminate opponents by hitting each other usingreplica weapons called airsoft guns The graph of a typicalsignal from the sensorized airsoft gun is shown in Figure 16A user can count the number of shots of a ball by attachingSenStick to his airsoft gun
45 Another Example Studentrsquos Unique Achievements andSurvey Results SenStick is also used in a lecture of IoT
Table 8 Confusion matrix of posture type recognition
PredictedC1 C2 C3 C4
ActualC1 good posture 1000 0000 0000 0000C2 bad posture 1 0000 0994 0006 0001C3 bad posture 2 0000 0006 0994 0000C4 bad posture 3 0000 0001 0000 0999
prototyping Students realize their unique IoT devices easilyby using SenSticks Here we introduce some of the developedIoT devices and survey results about the usefulness ofSenStick answered by students
Figure 17 shows a sensorized pencil with two SenSticksThe sensorized pencil is developed to recognize the charac-ters written by this pencil Even though two SenSticks areembedded to detect precise movement of pencil the size ofthe pencil is reasonable
Figure 18 shows a sensorized name tag anddigital signageSince we changed the battery from 60mAh to 900mAh thetag keeps running more than one week It aims to recordemployeesrsquo or studentsrsquo daily activities to support their healthThe digital signage informs the user who wears a sensorizedname tag about the userrsquos activity data when the user standsin front of it
Figure 19 shows a sensorized spoon used in the digitalfish catching game for children In this game the sensorizedspoon detects the scooping movement of the hand Thuschildren can catch the digital fish displayed by the projectionmapping by using sensorized spoonThis systemwas actuallyused at the universityrsquos open campus event and around100 children enjoyed this game and showed interest in IoTtechnology
The survey results obtained from the ten graduate stu-dents (ninemales and one female) who developed prototypesas mentioned above are shown in Table 9 A questionnaireis a five-grade form (5 strongly agree 4 agree a little 3neither agree nor disagree 2 disagree a little and 1 stronglydisagree) From the results of questions 1 and 2 manyparticipants think that SenStick platform is useful for IoT
14 Journal of Sensors
Position
Distance illumination model
Using trilateration method
a
Light 3
Light 2
Light 1
L = minus421d + 1496
L = minus063d + 501
05 1 15 2 25 3 35 4 450Distance between the illuminance sensor and the lighting device (m)
02468
1012141618
Illum
inan
ce v
alue
(lux
)
Distanceu
d1
d2
d3
l1 p1 x1
l3 p3 x3
l2 p2 x2
Figure 14 Indoor localization by sensorized glasses
Sensorized airsoft gun
Figure 15 Sensorized airsoft gun
research and IoT education Also the results of question 3show that many students are interested in the SenStick andwant to use it again From the results of questions 4 and 5we can see that usability and specification of SenStick arerelatively high There were opinions about the usefulness ofSenStick platform such as ldquoI could quickly try out ideas of IoTprototypingrdquo ldquoIt was easy to collect sensing datardquo and ldquoSinceSenStick is small it was easily attached tomost of the thingsrdquoOn the other hand there were opinions about improvements
such as ldquoI want to use higher sensing frequency (50ndash100Hz)rdquoldquoI want to process a sensing data on SenStickrdquo and ldquoI want tomodify the firmware itself depending on the application ofthe IoT devicerdquo
5 Conclusion
In this paper we have introduced a smart and comprehensivesensing platform called SenStick which is composed of
Journal of Sensors 15
Typical signal of airso gun
Hold a gun Ready
Shoot
1 2 3 4 5 6 70
00
30
60
91
21
51
82
12
42
73
03
33
63
94
24
54
85
15
45
76
06
36
66
97
27
57
88
18
48
79
09
39
69
910
210
510
811
111
411
712
012
312
612
913
2
minus600
minus400
minus200
0
200
400
600
800
acc_xacc_yacc_z
gyro_xgyro_y
acc_synthesis gyro_zgyro_synthesis
Figure 16 Time series of sensing data in shooting an airsoft gun
Sensorized pencil
Figure 17 Sensorized pencil
Sensorized name tagVisualization of sensor data
Figure 18 Sensorized name tag
Sensorized spoon Digital fishing game
Figure 19 Sensorized spoon
Table 9 Questionnaire results about SenStick
Question contents ScoreQ1 Do you think that SenStick will be useful for IoTresearch 49
Q2 Do you think that SenStick will be useful for IoTeducation 47
Q3 Do you want to use SenStick again 49Q4 Do you think that SenStick was easy to use 41Q5 Are you satisfied with specification of SenStick 41
hardware software and 3D case data The main purpose ofour platform is to simplify the sensorization of our personalbelongings
The hardware of SenStick is maybe one of the worldrsquossmallest sensor boards which includes 8 kinds of sensorsand BLE In addition to the hardware we also developedsupport software for managing monitoring recording andviewing the hardware By providing 3D case data all theresearchers can measure every tiny activity in daily life easilyand efficiently For an educational use it is also useful forunderstanding sensors and prototyping the desired systemincluding tiny censored things
Conflicts of Interest
The authors declare that there are no conflicts of interestregarding the publication of this paper
Acknowledgments
This work is supported in part by JST PRESTO and JSPSKAKENHI (Grant no 17J10021) and ldquoResearch and Devel-opment of Innovative Network Technologies to Create theFuturerdquo the Commissioned Research of National Instituteof Information and Communications Technology (NICT)Japan
16 Journal of Sensors
References
[1] O Brdiczka M Langet J Maisonnasse and J L CrowleyldquoDetecting human behavior models from multimodal observa-tion in a smart homerdquo IEEE Transactions on Automation Scienceand Engineering vol 6 no 4 pp 588ndash597 2009
[2] J Chen A Kam J Zhang N Liu and L Shue ldquoBathroomactivity monitoring based on soundrdquo in Pervasive ComputingThird International Conference PERVASIVE 2005 MunichGermanyMay 8-13 2005 Proceedings vol 3468 ofLectureNotesin Computer Science pp 47ndash61 Springer Berlin Germany2002
[3] G Bauer K Stockinger and P Lukowicz ldquoRecognizing theUse-Mode of Kitchen Appliances from Their Current Consump-tionrdquo in Smart Sensing and Context vol 5741 of Lecture Notesin Computer Science pp 163ndash176 Springer Berlin HeidelbergBerlin Heidelberg 2009
[4] K Ouchi and M Doi ldquoIndoor-outdoor activity recognition bya smartphonerdquo in Proceedings of the the 2012 ACM Conferencep 537 Pittsburgh Pennsylvania September 2012
[5] A Anjum and M U Ilyas ldquoActivity recognition using smart-phone sensorsrdquo in Proceedings of the IEEE 10th ConsumerCommunications and Networking Conference (CCNC rsquo13) pp914ndash919 IEEE January 2013
[6] O D Lara and M A Labrador ldquoA survey on human activityrecognition using wearable sensorsrdquo IEEE CommunicationsSurveys amp Tutorials vol 15 no 3 pp 1192ndash1209 2013
[7] ldquoJins memerdquo httpsjins-memecom[8] Y Matsuda T Ishioka T Hasegawa et al ldquoWaistonBelt A belt
for monitoring your real abdominal circumference foreverrdquo inProceedings of the ACM International Joint Conference on Perva-sive and Ubiquitous Computing and the 2015 ACM InternationalSymposium on Wearable Computers UbiComp and ISWC 2015pp 329ndash332 September 2015
[9] M Kurahashi K Murao T Terada and M Tsukamoto ldquoPer-sonal identification system based on rotation of toilet paperrollsrdquo in Proceedings of the IEEE International Conference onPervasive Computing and CommunicationsWorkshops (PerComWorkshops rsquo17) pp 521ndash526 IEEE Kona Hawaii USA March2017
[10] ldquoHapiforkrdquo httpswwwhapicom[11] Y Arakawa ldquoSenStick Sensorize every thingsrdquo in Proceedings of
the ACM International Joint Conference on Pervasive and Ubiq-uitous Computing and the 2015 ACM International SymposiumonWearable Computers UbiComp and ISWC 2015 pp 349ndash352September 2015
[12] Y Nakamura T Kanehira Y Arakawa and K YasumotoldquoSenStick 2 Ultra tiny all-in-one sensor with wireless chargingrdquoinProceedings of the 2016ACM International Joint Conference onPervasive and Ubiquitous Computing UbiComp 2016 pp 337ndash340 September 2016
[13] ldquoSensortag cc2541 (texas instruments)rdquo httpwwwticomtoolcc2541dk-sensor
[14] ldquoIot smart modulerdquo httpwwwalpscomjiotsmart[15] ldquoTsnd121 (atr-promotions)rdquo httpwwwatr-pcomproducts
TSND121html[16] ldquoWax9 (axivity)rdquo httpaxivitycomproductwax9[17] Uk biobank ldquoUk biobankrdquo httpwwwukbiobankacuk[18] J C Seidell and J Halberstadt ldquoThe global burden of obesity
and the challenges of preventionrdquo Annals of Nutrition andMetabolism vol 66 supplement 2 pp 7ndash12 2015
[19] ldquoWeltThe smart belt for fashion and healthrdquo httpswwwwelt-corpcom
[20] ldquoBelty power The belt your really deserverdquo httpswwwbeltyco
[21] H Nam J-H Kim and J-I Kim ldquoSmart Belt A wearabledevice for managing abdominal obesityrdquo in Proceedings of theInternational Conference on Big Data and Smart ComputingBigComp 2016 pp 430ndash434 January 2016
[22] K Moriya M Fujimoto Y Arakawa and K Yasumoto ldquoIndoorlocalization based on distance-illuminance model and activecontrol of lighting devicesrdquo in Proceedings of the 2016 Interna-tional Conference on Indoor Positioning and Indoor NavigationIPIN 2016 pp 1ndash6 October 2016
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal of
Volume 201
Submit your manuscripts athttpswwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
4 Journal of Sensors
Table1Com
paris
onof
competitives
ensors
SenS
tick
3Se
nStic
k 2
Sens
orTa
g (C
C265
0)Io
T Sm
art M
odul
eTS
ND
121
WA
X9
Size
(mm)
50(119882
)times10
(119867)times
10(119863
)50
(119882)times
10(119867
)times10
(119863)
42(119882
)times32
(119867)times
10(119863
)44
(119882)times
27(119867
)times11(119863
)37
(119882)times
46(119867
)times12
(119863)23
(119882)times
325(119867
)times76
(119863)
Weight(g)
35
28mdash
227
Stand-alon
erecording
Itimes
timestimes
ISm
artpho
neapp
II
Itimes
timesDesktop
app
I998779
timesI
IBa
ttery
life
15(hou
r)1(year)
1(year)
6(hou
r)6(hou
r)Ba
ttery
type
Rechargeablerem
ovable
Rechargeable
Coinbatte
ryCoinbatte
ryRe
chargeable
Rechargeable
Network
BLE
BLE
BLE
Bluetooth20
BLE
Datap
rocessing
Itimes
timestimes
timestimes
I2CEx
tension
Itimes
timestimes
timestimes
Sensors
Acceleratio
nI
II
II
Gyro
II
timesI
IMagnetic
II
II
IAirpressure
II
II
ITemperature
II
Itimes
IHum
idity
II
Itimes
timesLight
II
Itimes
timesUV
Itimes
Itimes
timesMi c
timesI
timestimes
times
Journal of Sensors 5
Flash memory(back side)
50GG
60GB
Temperature amphumidity(SHT-20)
UV(VEML6070)
Pressure(LPS25HB)
Light(BH1780GLI)
Charger(BQ24090)
Micro USB
Li-Po battery
USBUART(CP2104)
9-axis(MPU-9250)
BLE(nRF51822)
Figure 2 Hardware of SenStick 2 (8 sensors BLE flash memory etc)
Table 2 Sensors embedded in SenStick 23
Type of sensor Model number Power consumption9-Axis MPU-9250 280 120583Asim37mAPressure LPS25HBTR 4 120583Asim25 120583ATemperaturehumidity SHT20 270 120583Asim330 120583AIlluminance BH1780GLI-E2 120120583Asim200 120583AUltraviolet (UV) VEML6070 100 120583Asim250 120583A
times 5mm (119863) and its weight is around 3 (g) including thebattery On this tiny board eight sensors (acceleration gyromagnetic light UV temperature humidity and pressure)flash memory (32Mbytes) BLE battery and charging circuitare densely embedded as shown in Figure 2 The modelnumber and power consumption of the sensors embedded inSenStick are listed in Table 2 All the sensors are popular chipsused worldwide and their performances are ensured by eachmanufacturer
The novelty of our hardware is the design of the sensorboard composed of the elaborated combination to satisfythe requirements presented in Section 2 This contributionbreaks down into the following aspects
(1) Combination of BLE and Flash Memory Almost allthe BLE-based devices equipped no large flash memoryto record sensing data because these devices are designedon the premise that sensing data are always recorded onsmartphones connected via BLE However the transmissionspeed of BLE is not enough for sending all the sensing dataof multiple sensors with the high sampling frequency As aresult some data will be dropped in the transmission Sincethe design of existing BLE-based devices is not appropriatefor research purpose we designed a new BLE-based sensingboard with a large flash memory which allows users torecord all the sensing data precisely Also it means that it canwork without holding a smartphone This ability is useful forsensing themotion of animals that cannot have a smartphone
(2) Selection of Powerful and Energy-Saving SoC Weselected Nordicrsquos nRF52832 SoC as SenStickrsquos microcon-troller Because the nRF52832 SoC is built around a 32-bit ARM Cortex-M4F CPU with 512 kB + 64 kB RAM itis extremely powerful Moreover the embedded 24GHztransceiver supports Bluetooth low energy and has extremelylow power consumption and it is the worldrsquos smallest classmodule Each sensor is connected via two I2C buses inde-pendently processing high-speed sensor (acceleration gyroandmagnetic) and low-speed sensor (light UV temperaturehumidity and pressure) taking into account the differencein bus occupancy time This makes it possible to realize highsampling with the motion sensors
(3) Application-Friendly Battery Type and Shape Since Sen-Stick is designed to bemounted on objects such as chopstickstoothbrushes and glasses it is desirable that the shape is thinand compact However sensors such as SensorTag equippedwith readily available coin batteries are inappropriate inshape because the footprint is square Also in the researchnonrechargeable coin batteries are costly and troublesomeFor those reasons we selected a thin and compact lithiumpolymer battery on SenStick This type of battery was harderto obtain than expected Therefore the procurement ofbattery has become the bottleneck of development
SenStickrsquos firmware is newly designed and developed byconsidering aforementioned requirements Generic Attribute(GATT) profile of SenStick provided by firmware is shownin Figure 3 The GATT server in SenStick provides threeservices (SenStick Control Service Metadata Read Serviceand Sensor Service) The SenStick Control Service providesoperation instructions such as starting and stopping sensingand logging The metadata reading service provides thefunction of reading metainformation of log data The SensorService provides functions to specify the sensing and loggingoperation of each sensor (0 acceleration 1 gyro 2 magnetic3 light 4 UV 5 temperaturehumidity 6 pressure) Also itprovides real-time sensing data reading function and log datareading function
6 Journal of Sensors
(32-)
Peripheral
GATT server
SenStick Control Service
Metadata Read Service
Read-write
Data notication
Flashmemory
(0 acceleration 1 gyro 2 magnetic 3 light
4 UV 5 temperaturehumidity 6 pressure)
GATT client
Central
~Sensor Service (0 6)
Figure 3 Generic Attribute (GATT) profile of SenStick
Table 3 BLE communication specification
Type of content SpecAdvertising packet Complete UUID flagScan response Local name (SenStick)Advertising packet interval 100ms (30 s)rarr1000ms (150 s)rarrsleepConnection interval max 80msmin 20msSupervision timeout 4000msSlave latency 0
Basically all the sensing data will be notified to a smart-phone in real time up to a transmission capacity of BLE Ifthe amount of sensing data exceeds the transmission capacitya part of sensing data will be dropped Therefore the real-time sensing data reading function is used formonitoring thestatus or for an interactive application that does not requireprecise data For precise data acquisition the log data readingfunction which guarantees the data reading order is usedThe communication specifications of BLE are described inTable 3
As shown in Table 4 the fixed capacities of a flashmemory are allocated to each sensorThe number of samplesindicates the total number of samples that can be recorded bythe allocated capacities
SenStick stops logging when the number of samples ofone sensor exceeds the storage capacity allocated for it Themaximum recording time of acceleration gyroscope andmagnetic sensor is 4 hours 40 minutes at sampling intervalof 10 milliseconds (sampling frequency 100Hz) and 14 hours10 minutes at sampling period of 30 milliseconds (samplingfrequency 33Hz) The maximum recording time of othersensors is 9 hours 26 minutes at sampling period of 200milliseconds (sampling frequency 5Hz)
32 Software We provide mobile applications for iOS andAndroid and a library for nodejs platform Common func-tions of them are operating function parameter settingfunction and real-time data view function An additional
Table 4 Allocated storage capacity per sensor
Amount of allocatedstorage capacity Samples
Acceleration 102Mbytes 17M samplesGyro 102Mbytes 17M samplesMagnetic 102Mbytes 17M samplesIlluminance 340KB 170K samplesUV (UV) 340KB 170K samplesHumidity 170KB 170K samplesAir temperature 170KB 170K samples
function of iOS application is a firmware update functionThose of Android application and nodejs applications areground truth video recording function and an ability to con-nect multiple SenSticks respectively Since all these softwareapplications are shared on GitHub as an own source softwareall the users can create their application based on their ownpurposes
Communication Protocol Here we explain the communica-tion protocol between SenStick and softwareThe explanationis based on ldquonode-SenStickrdquo which is our developed libraryfor nodejs Figure 4 shows a protocol sequence betweenSenStick and the central terminal (node-SenStick on Mac)
(I) Discovery First of all the application needs to find thesurrounding SenStick by using SenStickdiscovery() methodin our library When SenStickdiscovery() method finds aSenStick a callback function with an instance of SenStickobject containing the SenStickrsquos UUID and LocalName asarguments is called
(II) Connect If the target SenStick is found by the discoveryprocess the application is going to establish a connec-tion with it by calling SenStickprototypeconnectAndSetUp()method After establishing the connection this methodacquires the information necessary for operation from thetarget SenStick
Journal of Sensors 7
Central
Node-SenStick SenStick
SenStickdiscovery()
SenStickprototypeconnectAndSetUp()
Establish connection
SenStickprototypewriteSensorMeasurementCong()
Reect setting
SenStickprototypestartSensingAndLogging()
SenStickprototypenotifySensor()Start sensing
Select sensor to notify
Sensing data
SenStickprototypestopSensingAndLogging()
Stop sensing
Enable log notication
Select log data to notify
SenStickprototypewriteSensorLogReadoutTargetID()
Logging data
SenStickprototypedisconnect()
Disconnect
SenStickprototypenotifySensorLogData()
Periperal
(I) Discovery
(II) Connect
(III) Setting
(V) Obtaining log
(VI) Disconnect
(IV) Start sensingamp
RT receiving
Figure 4 Communication sequence between SenStick and node-SenStick
(III) Setting After establishing the connection the appli-cation sets three parameters (operation mode samplinginterval and measurement range) for each sensor by usingSenStickprototypewriteSensorMeasurementConfig()methodThe operation mode is a parameter that sets valida-tioninvalidation of sensing and logging of each sensor Thesampling interval and measurement range are parameters ofeach sensorwhichwill be changed by the sensing purpose andapplications
(IV) Start Sensing and Real-Time Data Receiving When Sen-StickprototypestartSensingAndLogging() method is calledSenStick starts sensing and logging Only valid sensorsrsquodata are transmitted via BLE and also recorded into aflash memory ID of each log is automatically assignedDuring sensing SenStick notifies sensing data in realtime To receive this notification the application callsSenStickprototypenotifySensor() method (Sensor representseach sensor name) and activates the notification for
8 Journal of Sensors
(i) Frequency(ii) Range
Device select Main view Settings Firmware update
Real-time view Log list in SenStick Download from SenStick
Selected sensorSensor name
Sensing frequency
Range of y-axis
Figure 5 iOS application for SenStick
the sensor When a notification of sensing data isreceived sensorChange event will be issued If Sen-StickprototypestopSensingAndLogging() method is calledSenStick stops sensing and logging
(V) Log Data Reading To read the log data recordedin the flash memory to the application the applicationcalls SenStickprototypenotifySensorLogData()method (Sen-sor represents each sensor name) and enables log datanotification If the application specifies a valid log ID bycalling SenStickprototypewriteSensorLogReadoutTargetID()method SenStick starts informing the sensing data recordedin the log sequentially To receive notification of log datasensorLogDataReceived event will be issued By receiving thisnotification by the application it is possible to acquire the logdata recorded in the memory
(VI) Disconnect After finishing the data acquisition Sen-Stickprototypedisconnect() method will be called to releasethe connection of SenStick And SenStick will begin advertis-ing again
User InterfaceWe explain the user interface of the applicationbased on iOS version Figure 5 shows a sophisticated mobileapplication for SenStick As shown in Figure 5 the userselects target SenStick from devices listThen the applicationshows the main view In the main view the user can selectsensors to be acquired For immediate data monitoring allthe received data via BLE can be shown in the graph view inreal time The application can control all the settings of eachsensor such as sample frequency and sensing range And alsoit has firmware update function Sensing data stored in theflash memory of SenStick is viewed from log list and it can be
downloaded to the smartphone in CSV formatThis functionprevents data loss caused by BLE communication
33 3D Case Data In order to make sensorized things anattachment case for mounting the SenStick is required Wedesigned the basic 3D case of SenStick By extending thisbasic case users can design derivative cases for chopsticksglasses and so on Actually our laboratory has succeeded indesigning cases for several things as shown in Figure 6 Allof these cases data are shared on our community site Eachcase is designed in a few minutes thanks to the basic caseFor example SenStick case for glasses is made by combiningbasic case of SenStick and open 3D data of glasses distributedby i Design Studio In addition due to the popularizationof 3D printers in the world and enhancement of tools andreferences for 3D modeling all the users can print out theirdesired SenStick cases by themselves and attach our sensor toeverything
4 Case Studies of SenStick
SenStick an ultra small sensor can be embedded into varioustiny things in our daily life Our first target is our dailybelongings such as glasses chopsticks and pens becausesensorization of daily belongings is good to intuitively under-standwhat can be sensed andwhich sensor is usefuleffectiveSenStick can upgrade normal glasses to intelligent oneswhichcan monitor a head movement posture and the amount ofsunlight received If SenStick is embedded into chopsticksand pens we can observe eating activities and handwritingactivities respectively In the future we will embed it into astick for elderlymonitoring sports gearsmusical instrument
Journal of Sensors 9
Chopsticks Tooth brush Glasses Umbrella Name card Cookware
A user can design original case data in a few minutes
Basic 3D case of SenStick
Figure 6 3D case data of SenStick
Sensorized
Chopsticks
Figure 7 Eating activity recognition by sensorized chopsticks
kitchen utensils animals stuffed toy a remote control andany kind of things that can be sensing targets In this sectionwe introduce some case studies results utilizing SenStick
41 Eating Behavior Recognition by Sensorized Chopsticks Inmany cases East Asian people eat meals with chopsticksBy analyzing the data obtained from chopsticks there is apossibility that we can estimate the eating habits includinganswers for questions like ldquohow did you eatrdquo and ldquowhatdid you eatrdquo and so forth These data are very useful formaking healthcare applications and leading a healthy lifeThis is one reason why we develop the sensorized chopsticksusing SenStick shown in Figure 7
Research goals of the eating behavior recognition by sen-sorized chopsticks are as follows G1 (behavior) recognizingthe timing and speed and detecting ldquocroprdquo and ldquointakerdquoaction G2 (food type) classifying food content like ricenoodles and so forth from the eating behavior and G3(person) identifying an individual from using chopsticks Inthis paper we show the research results on G1 and G2
We conducted basic experiments to recognize the eatingbehavior (G1 and G2) We used acceleration and gyroscopedata measured by SenStick where the sensing rate is 10HzThe purpose of experiment 1 is to understand the basic
characteristics of sensing data from sensorized chopstickwhen a participant ate chips The purpose of experiment 2 isto identify the food type from the difference in movement ofchopsticks We assumed 4 classes (C1 rice and dish C2 misosoup C3 noodles and C4 shrimp) as food types
Figure 8 shows time series of sensing data for a set ofeating actions in experiment 1 The eating actions are cropaction and intake action In crop action 6-axis motionsensors detect movement (open and close) of the chopsticksIn intake action sensors detect movement of the arm fromthe dish to the mouth Each graph in Figure 8 shows certaincorrelation with both actions
Figure 9 shows a typical signal for each food in exper-iment 2 In this experiment we calculate the recognitionaccuracy of food type classification by using Random Forestclassier ofWeka Tables 5 and 6 show the classification resultsThe average recognition accuracy was 896 If we did nothave a SenStick it would be hard to develop the sensorizedchopsticks that can be used in the actual environment
42 Daily Activity Monitoring by a Sensorized Belt Recentlycorpulent people are rapidly increasing and about 700 mil-lion people belong to the obesity group [18] Nowadays manyoffice workers use computers heavily and their sitting time
10 Journal of Sensors
Step 1 crop action Step 2 intake action
Lower
Upper
2 4 6 8 10 12 140Elapsed time (sec)
2 4 6 8 10 12 140Elapsed time (sec)
2 4 6 8 10 12 140Elapsed time (sec)
2 4 6 8 10 12 140Elapsed time (sec)
X
Y
Z
Synthetic
X
Y
Z
Synthetic
X
Y
Z
Synthetic
X
Y
Z
Synthetic
minus250
minus150
minus50
50
150
250
Ang
ular
velo
city
(rad
s)
minus20
minus10
01020
Acce
lera
tion
(ms
2)
minus250
minus150
minus50
50
150
250
Ang
ular
velo
city
(rad
s)
minus20
minus10
01020
Acce
lera
tion
(ms
2)
Open1 Close2 Raise3 Eat4 Drop5
1 2 3 4 5
Figure 8 Time series of sensing data in eating activity
in office increases Keeping incorrect postures during longworking time increases the risk of obesity of office workersTherefore it is necessary tomonitor the daily activity of officeworkers for correcting the bad habits
Motivated by this we have developed a sensorized beltcalled Waiston Belt [8] which aims to monitor a userrsquosdaily activities for maintaining health As products of belt-type wearable devices Welt [19] and Belty [20] are on saleThese provide the function of waist measurement activity
tracking and step counting However API for obtainingsensing data is not provided and it is not suitable for researchpurposes Generally it is hard for ordinary researchers toembed sensors battery BLE and vibrator into a small buckle-size case that can be worn in everyday life Nam et al [21]have proposed a belt-attached device for monitoring postureHowever the device has not reached a sufficient level to beused on a daily basis because of exposed wiring and uncoolappearance However by extending SenStick we succeeded
Journal of Sensors 11
C1
C2 C3
C4
Typical signal of each food
C1 rice amp dish C2 miso soup C3 noodles C4 shrimp
400
200
0
minus200
minus400
Figure 9 Eating activity recognition by sensorized chopsticks
SenStick
Waiston Belt
Figure 10 Waiston Belt based on SenStick
Table 5 Experimental result (food type)
Food type Number ofpieces of data Precision Recall 119865-Measure
C1 rice and dish 3011 0894 0945 0919C2 miso soup 2187 0907 0841 0873C3 noodles 2761 088 0896 0888C4 shrimp 2091 0907 088 0893
0896 0896 0895
to develop a stylish prototype of Waiston Belt shown inFigure 10 quickly Since the battery life of SenStick is 15 hourswith a small battery (60mAh) we can sense continuously inour daily lives As shown in Figure 11 it is possible to estimatevarious usersrsquo contexts (How much did you walk Are youusing the stairs Is posture good at work) by analyzing timeseries of sensing data obtained from the sensorized belt
An important point in this use case is that usersrsquo contextscan be estimated by combining multiple sensing data For
Table 6 Confusion matrix of food type recognition
PredictedC1 C2 C3 C4
ActualC1 rice and dish 094 004 001 000C2 miso soup 012 084 004 000C3 noodles 002 002 090 006C4 shrimp 001 001 011 088
example climbing stairs is estimated from the changes ofacceleration and atmospheric pressure The change of thehumidity data will be helpful to detect that a user enteredan air-conditioned room A SenStick makes it possible toeasily carry out research on behavior recognition with acombination of such multiple sensors
We introduce basic experiments about the posture recog-nition during deskwork by using this beltThe purpose of thisexperiment is to identify the posture type from the attitude
12 Journal of Sensors
Enter the lab
HumidityTemp
Humidity dropping
Pressure dropping
Working
Walking WalkingClimbing stairs Sitting (good posture) Sitting (bad posture)
AccsXAccsYAccsZ
60
40
20
1003
1002
1001
1000
999
25
2
15
1
05
0
minus15
minus1
minus05
Air pressure
Figure 11 Time series of sensing data in daily activity
C1 good posture C2 bad posture 1
C3 bad posture 2 C4 bad posture 3
Figure 12 Time series of sensing data in posture monitoring
angle of the belt calculated from 3 axial acceleration sensorsWe set the sensing rate as 10Hz and collected 6000 samples(10min) for each posture We assumed 4 classes (C1 goodposture C2 bad posture 1 C3 bad posture 2 and C4 badposture 3) as shown in Figure 12
In this experiment we calculated the recognition accu-racy of posture type classification by using Random Forestclassifier of Weka Tables 7 and 8 show the classificationresults The average recognition accuracy was 997 Inthe future based on the result of posture estimation we
Journal of Sensors 13
Sensorized glasses
Figure 13 Sensorized glasses
Table 7 Experimental result (posture type)
Posture type Number ofpieces of data Precision Recall 119865-Measure
C1 good posture 6000 1000 1000 1000C2 bad posture 1 6000 0994 0994 0994C3 bad posture 2 6000 0994 0994 0994C4 bad posture 3 6000 0999 0999 0999
0997 0997 0997
will develop a software application that suggests correctinghisher bad posture to a right posture during desk work
43 Indoor Localization by Sensorized Glasses We have pro-posed an indoor localization system using illuminance-basedtrilateration method [22] In this system we assume that theperson wears glasses having an illuminance sensor as shownin Figure 13 and controllable light sources are deployed in thetarget environment Our system controls lights andmeasuresthe variation of illuminance at the glasses As shown inFigure 14 the distances (1198891ndash1198893) from each light (1198971ndash1198973) to theuser (119906) are estimated from the distance-illumination modelconstructed in advance and userrsquos position is estimated basedon trilateration method using estimated distances
As a result of an experiment where three controllablelight sources were deployed in a room we confirmed thatour system could estimate the position within 05m erroron average By using SenStick we could easily develop thesensorized glasses equipped with illuminance sensor
44 Sensorized Airsoft Gun We have developed the sen-sorized airsoft gun for activity recognition in airsoft gameas shown in Figure 15 Airsoft game is a sport in whichparticipants eliminate opponents by hitting each other usingreplica weapons called airsoft guns The graph of a typicalsignal from the sensorized airsoft gun is shown in Figure 16A user can count the number of shots of a ball by attachingSenStick to his airsoft gun
45 Another Example Studentrsquos Unique Achievements andSurvey Results SenStick is also used in a lecture of IoT
Table 8 Confusion matrix of posture type recognition
PredictedC1 C2 C3 C4
ActualC1 good posture 1000 0000 0000 0000C2 bad posture 1 0000 0994 0006 0001C3 bad posture 2 0000 0006 0994 0000C4 bad posture 3 0000 0001 0000 0999
prototyping Students realize their unique IoT devices easilyby using SenSticks Here we introduce some of the developedIoT devices and survey results about the usefulness ofSenStick answered by students
Figure 17 shows a sensorized pencil with two SenSticksThe sensorized pencil is developed to recognize the charac-ters written by this pencil Even though two SenSticks areembedded to detect precise movement of pencil the size ofthe pencil is reasonable
Figure 18 shows a sensorized name tag anddigital signageSince we changed the battery from 60mAh to 900mAh thetag keeps running more than one week It aims to recordemployeesrsquo or studentsrsquo daily activities to support their healthThe digital signage informs the user who wears a sensorizedname tag about the userrsquos activity data when the user standsin front of it
Figure 19 shows a sensorized spoon used in the digitalfish catching game for children In this game the sensorizedspoon detects the scooping movement of the hand Thuschildren can catch the digital fish displayed by the projectionmapping by using sensorized spoonThis systemwas actuallyused at the universityrsquos open campus event and around100 children enjoyed this game and showed interest in IoTtechnology
The survey results obtained from the ten graduate stu-dents (ninemales and one female) who developed prototypesas mentioned above are shown in Table 9 A questionnaireis a five-grade form (5 strongly agree 4 agree a little 3neither agree nor disagree 2 disagree a little and 1 stronglydisagree) From the results of questions 1 and 2 manyparticipants think that SenStick platform is useful for IoT
14 Journal of Sensors
Position
Distance illumination model
Using trilateration method
a
Light 3
Light 2
Light 1
L = minus421d + 1496
L = minus063d + 501
05 1 15 2 25 3 35 4 450Distance between the illuminance sensor and the lighting device (m)
02468
1012141618
Illum
inan
ce v
alue
(lux
)
Distanceu
d1
d2
d3
l1 p1 x1
l3 p3 x3
l2 p2 x2
Figure 14 Indoor localization by sensorized glasses
Sensorized airsoft gun
Figure 15 Sensorized airsoft gun
research and IoT education Also the results of question 3show that many students are interested in the SenStick andwant to use it again From the results of questions 4 and 5we can see that usability and specification of SenStick arerelatively high There were opinions about the usefulness ofSenStick platform such as ldquoI could quickly try out ideas of IoTprototypingrdquo ldquoIt was easy to collect sensing datardquo and ldquoSinceSenStick is small it was easily attached tomost of the thingsrdquoOn the other hand there were opinions about improvements
such as ldquoI want to use higher sensing frequency (50ndash100Hz)rdquoldquoI want to process a sensing data on SenStickrdquo and ldquoI want tomodify the firmware itself depending on the application ofthe IoT devicerdquo
5 Conclusion
In this paper we have introduced a smart and comprehensivesensing platform called SenStick which is composed of
Journal of Sensors 15
Typical signal of airso gun
Hold a gun Ready
Shoot
1 2 3 4 5 6 70
00
30
60
91
21
51
82
12
42
73
03
33
63
94
24
54
85
15
45
76
06
36
66
97
27
57
88
18
48
79
09
39
69
910
210
510
811
111
411
712
012
312
612
913
2
minus600
minus400
minus200
0
200
400
600
800
acc_xacc_yacc_z
gyro_xgyro_y
acc_synthesis gyro_zgyro_synthesis
Figure 16 Time series of sensing data in shooting an airsoft gun
Sensorized pencil
Figure 17 Sensorized pencil
Sensorized name tagVisualization of sensor data
Figure 18 Sensorized name tag
Sensorized spoon Digital fishing game
Figure 19 Sensorized spoon
Table 9 Questionnaire results about SenStick
Question contents ScoreQ1 Do you think that SenStick will be useful for IoTresearch 49
Q2 Do you think that SenStick will be useful for IoTeducation 47
Q3 Do you want to use SenStick again 49Q4 Do you think that SenStick was easy to use 41Q5 Are you satisfied with specification of SenStick 41
hardware software and 3D case data The main purpose ofour platform is to simplify the sensorization of our personalbelongings
The hardware of SenStick is maybe one of the worldrsquossmallest sensor boards which includes 8 kinds of sensorsand BLE In addition to the hardware we also developedsupport software for managing monitoring recording andviewing the hardware By providing 3D case data all theresearchers can measure every tiny activity in daily life easilyand efficiently For an educational use it is also useful forunderstanding sensors and prototyping the desired systemincluding tiny censored things
Conflicts of Interest
The authors declare that there are no conflicts of interestregarding the publication of this paper
Acknowledgments
This work is supported in part by JST PRESTO and JSPSKAKENHI (Grant no 17J10021) and ldquoResearch and Devel-opment of Innovative Network Technologies to Create theFuturerdquo the Commissioned Research of National Instituteof Information and Communications Technology (NICT)Japan
16 Journal of Sensors
References
[1] O Brdiczka M Langet J Maisonnasse and J L CrowleyldquoDetecting human behavior models from multimodal observa-tion in a smart homerdquo IEEE Transactions on Automation Scienceand Engineering vol 6 no 4 pp 588ndash597 2009
[2] J Chen A Kam J Zhang N Liu and L Shue ldquoBathroomactivity monitoring based on soundrdquo in Pervasive ComputingThird International Conference PERVASIVE 2005 MunichGermanyMay 8-13 2005 Proceedings vol 3468 ofLectureNotesin Computer Science pp 47ndash61 Springer Berlin Germany2002
[3] G Bauer K Stockinger and P Lukowicz ldquoRecognizing theUse-Mode of Kitchen Appliances from Their Current Consump-tionrdquo in Smart Sensing and Context vol 5741 of Lecture Notesin Computer Science pp 163ndash176 Springer Berlin HeidelbergBerlin Heidelberg 2009
[4] K Ouchi and M Doi ldquoIndoor-outdoor activity recognition bya smartphonerdquo in Proceedings of the the 2012 ACM Conferencep 537 Pittsburgh Pennsylvania September 2012
[5] A Anjum and M U Ilyas ldquoActivity recognition using smart-phone sensorsrdquo in Proceedings of the IEEE 10th ConsumerCommunications and Networking Conference (CCNC rsquo13) pp914ndash919 IEEE January 2013
[6] O D Lara and M A Labrador ldquoA survey on human activityrecognition using wearable sensorsrdquo IEEE CommunicationsSurveys amp Tutorials vol 15 no 3 pp 1192ndash1209 2013
[7] ldquoJins memerdquo httpsjins-memecom[8] Y Matsuda T Ishioka T Hasegawa et al ldquoWaistonBelt A belt
for monitoring your real abdominal circumference foreverrdquo inProceedings of the ACM International Joint Conference on Perva-sive and Ubiquitous Computing and the 2015 ACM InternationalSymposium on Wearable Computers UbiComp and ISWC 2015pp 329ndash332 September 2015
[9] M Kurahashi K Murao T Terada and M Tsukamoto ldquoPer-sonal identification system based on rotation of toilet paperrollsrdquo in Proceedings of the IEEE International Conference onPervasive Computing and CommunicationsWorkshops (PerComWorkshops rsquo17) pp 521ndash526 IEEE Kona Hawaii USA March2017
[10] ldquoHapiforkrdquo httpswwwhapicom[11] Y Arakawa ldquoSenStick Sensorize every thingsrdquo in Proceedings of
the ACM International Joint Conference on Pervasive and Ubiq-uitous Computing and the 2015 ACM International SymposiumonWearable Computers UbiComp and ISWC 2015 pp 349ndash352September 2015
[12] Y Nakamura T Kanehira Y Arakawa and K YasumotoldquoSenStick 2 Ultra tiny all-in-one sensor with wireless chargingrdquoinProceedings of the 2016ACM International Joint Conference onPervasive and Ubiquitous Computing UbiComp 2016 pp 337ndash340 September 2016
[13] ldquoSensortag cc2541 (texas instruments)rdquo httpwwwticomtoolcc2541dk-sensor
[14] ldquoIot smart modulerdquo httpwwwalpscomjiotsmart[15] ldquoTsnd121 (atr-promotions)rdquo httpwwwatr-pcomproducts
TSND121html[16] ldquoWax9 (axivity)rdquo httpaxivitycomproductwax9[17] Uk biobank ldquoUk biobankrdquo httpwwwukbiobankacuk[18] J C Seidell and J Halberstadt ldquoThe global burden of obesity
and the challenges of preventionrdquo Annals of Nutrition andMetabolism vol 66 supplement 2 pp 7ndash12 2015
[19] ldquoWeltThe smart belt for fashion and healthrdquo httpswwwwelt-corpcom
[20] ldquoBelty power The belt your really deserverdquo httpswwwbeltyco
[21] H Nam J-H Kim and J-I Kim ldquoSmart Belt A wearabledevice for managing abdominal obesityrdquo in Proceedings of theInternational Conference on Big Data and Smart ComputingBigComp 2016 pp 430ndash434 January 2016
[22] K Moriya M Fujimoto Y Arakawa and K Yasumoto ldquoIndoorlocalization based on distance-illuminance model and activecontrol of lighting devicesrdquo in Proceedings of the 2016 Interna-tional Conference on Indoor Positioning and Indoor NavigationIPIN 2016 pp 1ndash6 October 2016
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal of
Volume 201
Submit your manuscripts athttpswwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
Journal of Sensors 5
Flash memory(back side)
50GG
60GB
Temperature amphumidity(SHT-20)
UV(VEML6070)
Pressure(LPS25HB)
Light(BH1780GLI)
Charger(BQ24090)
Micro USB
Li-Po battery
USBUART(CP2104)
9-axis(MPU-9250)
BLE(nRF51822)
Figure 2 Hardware of SenStick 2 (8 sensors BLE flash memory etc)
Table 2 Sensors embedded in SenStick 23
Type of sensor Model number Power consumption9-Axis MPU-9250 280 120583Asim37mAPressure LPS25HBTR 4 120583Asim25 120583ATemperaturehumidity SHT20 270 120583Asim330 120583AIlluminance BH1780GLI-E2 120120583Asim200 120583AUltraviolet (UV) VEML6070 100 120583Asim250 120583A
times 5mm (119863) and its weight is around 3 (g) including thebattery On this tiny board eight sensors (acceleration gyromagnetic light UV temperature humidity and pressure)flash memory (32Mbytes) BLE battery and charging circuitare densely embedded as shown in Figure 2 The modelnumber and power consumption of the sensors embedded inSenStick are listed in Table 2 All the sensors are popular chipsused worldwide and their performances are ensured by eachmanufacturer
The novelty of our hardware is the design of the sensorboard composed of the elaborated combination to satisfythe requirements presented in Section 2 This contributionbreaks down into the following aspects
(1) Combination of BLE and Flash Memory Almost allthe BLE-based devices equipped no large flash memoryto record sensing data because these devices are designedon the premise that sensing data are always recorded onsmartphones connected via BLE However the transmissionspeed of BLE is not enough for sending all the sensing dataof multiple sensors with the high sampling frequency As aresult some data will be dropped in the transmission Sincethe design of existing BLE-based devices is not appropriatefor research purpose we designed a new BLE-based sensingboard with a large flash memory which allows users torecord all the sensing data precisely Also it means that it canwork without holding a smartphone This ability is useful forsensing themotion of animals that cannot have a smartphone
(2) Selection of Powerful and Energy-Saving SoC Weselected Nordicrsquos nRF52832 SoC as SenStickrsquos microcon-troller Because the nRF52832 SoC is built around a 32-bit ARM Cortex-M4F CPU with 512 kB + 64 kB RAM itis extremely powerful Moreover the embedded 24GHztransceiver supports Bluetooth low energy and has extremelylow power consumption and it is the worldrsquos smallest classmodule Each sensor is connected via two I2C buses inde-pendently processing high-speed sensor (acceleration gyroandmagnetic) and low-speed sensor (light UV temperaturehumidity and pressure) taking into account the differencein bus occupancy time This makes it possible to realize highsampling with the motion sensors
(3) Application-Friendly Battery Type and Shape Since Sen-Stick is designed to bemounted on objects such as chopstickstoothbrushes and glasses it is desirable that the shape is thinand compact However sensors such as SensorTag equippedwith readily available coin batteries are inappropriate inshape because the footprint is square Also in the researchnonrechargeable coin batteries are costly and troublesomeFor those reasons we selected a thin and compact lithiumpolymer battery on SenStick This type of battery was harderto obtain than expected Therefore the procurement ofbattery has become the bottleneck of development
SenStickrsquos firmware is newly designed and developed byconsidering aforementioned requirements Generic Attribute(GATT) profile of SenStick provided by firmware is shownin Figure 3 The GATT server in SenStick provides threeservices (SenStick Control Service Metadata Read Serviceand Sensor Service) The SenStick Control Service providesoperation instructions such as starting and stopping sensingand logging The metadata reading service provides thefunction of reading metainformation of log data The SensorService provides functions to specify the sensing and loggingoperation of each sensor (0 acceleration 1 gyro 2 magnetic3 light 4 UV 5 temperaturehumidity 6 pressure) Also itprovides real-time sensing data reading function and log datareading function
6 Journal of Sensors
(32-)
Peripheral
GATT server
SenStick Control Service
Metadata Read Service
Read-write
Data notication
Flashmemory
(0 acceleration 1 gyro 2 magnetic 3 light
4 UV 5 temperaturehumidity 6 pressure)
GATT client
Central
~Sensor Service (0 6)
Figure 3 Generic Attribute (GATT) profile of SenStick
Table 3 BLE communication specification
Type of content SpecAdvertising packet Complete UUID flagScan response Local name (SenStick)Advertising packet interval 100ms (30 s)rarr1000ms (150 s)rarrsleepConnection interval max 80msmin 20msSupervision timeout 4000msSlave latency 0
Basically all the sensing data will be notified to a smart-phone in real time up to a transmission capacity of BLE Ifthe amount of sensing data exceeds the transmission capacitya part of sensing data will be dropped Therefore the real-time sensing data reading function is used formonitoring thestatus or for an interactive application that does not requireprecise data For precise data acquisition the log data readingfunction which guarantees the data reading order is usedThe communication specifications of BLE are described inTable 3
As shown in Table 4 the fixed capacities of a flashmemory are allocated to each sensorThe number of samplesindicates the total number of samples that can be recorded bythe allocated capacities
SenStick stops logging when the number of samples ofone sensor exceeds the storage capacity allocated for it Themaximum recording time of acceleration gyroscope andmagnetic sensor is 4 hours 40 minutes at sampling intervalof 10 milliseconds (sampling frequency 100Hz) and 14 hours10 minutes at sampling period of 30 milliseconds (samplingfrequency 33Hz) The maximum recording time of othersensors is 9 hours 26 minutes at sampling period of 200milliseconds (sampling frequency 5Hz)
32 Software We provide mobile applications for iOS andAndroid and a library for nodejs platform Common func-tions of them are operating function parameter settingfunction and real-time data view function An additional
Table 4 Allocated storage capacity per sensor
Amount of allocatedstorage capacity Samples
Acceleration 102Mbytes 17M samplesGyro 102Mbytes 17M samplesMagnetic 102Mbytes 17M samplesIlluminance 340KB 170K samplesUV (UV) 340KB 170K samplesHumidity 170KB 170K samplesAir temperature 170KB 170K samples
function of iOS application is a firmware update functionThose of Android application and nodejs applications areground truth video recording function and an ability to con-nect multiple SenSticks respectively Since all these softwareapplications are shared on GitHub as an own source softwareall the users can create their application based on their ownpurposes
Communication Protocol Here we explain the communica-tion protocol between SenStick and softwareThe explanationis based on ldquonode-SenStickrdquo which is our developed libraryfor nodejs Figure 4 shows a protocol sequence betweenSenStick and the central terminal (node-SenStick on Mac)
(I) Discovery First of all the application needs to find thesurrounding SenStick by using SenStickdiscovery() methodin our library When SenStickdiscovery() method finds aSenStick a callback function with an instance of SenStickobject containing the SenStickrsquos UUID and LocalName asarguments is called
(II) Connect If the target SenStick is found by the discoveryprocess the application is going to establish a connec-tion with it by calling SenStickprototypeconnectAndSetUp()method After establishing the connection this methodacquires the information necessary for operation from thetarget SenStick
Journal of Sensors 7
Central
Node-SenStick SenStick
SenStickdiscovery()
SenStickprototypeconnectAndSetUp()
Establish connection
SenStickprototypewriteSensorMeasurementCong()
Reect setting
SenStickprototypestartSensingAndLogging()
SenStickprototypenotifySensor()Start sensing
Select sensor to notify
Sensing data
SenStickprototypestopSensingAndLogging()
Stop sensing
Enable log notication
Select log data to notify
SenStickprototypewriteSensorLogReadoutTargetID()
Logging data
SenStickprototypedisconnect()
Disconnect
SenStickprototypenotifySensorLogData()
Periperal
(I) Discovery
(II) Connect
(III) Setting
(V) Obtaining log
(VI) Disconnect
(IV) Start sensingamp
RT receiving
Figure 4 Communication sequence between SenStick and node-SenStick
(III) Setting After establishing the connection the appli-cation sets three parameters (operation mode samplinginterval and measurement range) for each sensor by usingSenStickprototypewriteSensorMeasurementConfig()methodThe operation mode is a parameter that sets valida-tioninvalidation of sensing and logging of each sensor Thesampling interval and measurement range are parameters ofeach sensorwhichwill be changed by the sensing purpose andapplications
(IV) Start Sensing and Real-Time Data Receiving When Sen-StickprototypestartSensingAndLogging() method is calledSenStick starts sensing and logging Only valid sensorsrsquodata are transmitted via BLE and also recorded into aflash memory ID of each log is automatically assignedDuring sensing SenStick notifies sensing data in realtime To receive this notification the application callsSenStickprototypenotifySensor() method (Sensor representseach sensor name) and activates the notification for
8 Journal of Sensors
(i) Frequency(ii) Range
Device select Main view Settings Firmware update
Real-time view Log list in SenStick Download from SenStick
Selected sensorSensor name
Sensing frequency
Range of y-axis
Figure 5 iOS application for SenStick
the sensor When a notification of sensing data isreceived sensorChange event will be issued If Sen-StickprototypestopSensingAndLogging() method is calledSenStick stops sensing and logging
(V) Log Data Reading To read the log data recordedin the flash memory to the application the applicationcalls SenStickprototypenotifySensorLogData()method (Sen-sor represents each sensor name) and enables log datanotification If the application specifies a valid log ID bycalling SenStickprototypewriteSensorLogReadoutTargetID()method SenStick starts informing the sensing data recordedin the log sequentially To receive notification of log datasensorLogDataReceived event will be issued By receiving thisnotification by the application it is possible to acquire the logdata recorded in the memory
(VI) Disconnect After finishing the data acquisition Sen-Stickprototypedisconnect() method will be called to releasethe connection of SenStick And SenStick will begin advertis-ing again
User InterfaceWe explain the user interface of the applicationbased on iOS version Figure 5 shows a sophisticated mobileapplication for SenStick As shown in Figure 5 the userselects target SenStick from devices listThen the applicationshows the main view In the main view the user can selectsensors to be acquired For immediate data monitoring allthe received data via BLE can be shown in the graph view inreal time The application can control all the settings of eachsensor such as sample frequency and sensing range And alsoit has firmware update function Sensing data stored in theflash memory of SenStick is viewed from log list and it can be
downloaded to the smartphone in CSV formatThis functionprevents data loss caused by BLE communication
33 3D Case Data In order to make sensorized things anattachment case for mounting the SenStick is required Wedesigned the basic 3D case of SenStick By extending thisbasic case users can design derivative cases for chopsticksglasses and so on Actually our laboratory has succeeded indesigning cases for several things as shown in Figure 6 Allof these cases data are shared on our community site Eachcase is designed in a few minutes thanks to the basic caseFor example SenStick case for glasses is made by combiningbasic case of SenStick and open 3D data of glasses distributedby i Design Studio In addition due to the popularizationof 3D printers in the world and enhancement of tools andreferences for 3D modeling all the users can print out theirdesired SenStick cases by themselves and attach our sensor toeverything
4 Case Studies of SenStick
SenStick an ultra small sensor can be embedded into varioustiny things in our daily life Our first target is our dailybelongings such as glasses chopsticks and pens becausesensorization of daily belongings is good to intuitively under-standwhat can be sensed andwhich sensor is usefuleffectiveSenStick can upgrade normal glasses to intelligent oneswhichcan monitor a head movement posture and the amount ofsunlight received If SenStick is embedded into chopsticksand pens we can observe eating activities and handwritingactivities respectively In the future we will embed it into astick for elderlymonitoring sports gearsmusical instrument
Journal of Sensors 9
Chopsticks Tooth brush Glasses Umbrella Name card Cookware
A user can design original case data in a few minutes
Basic 3D case of SenStick
Figure 6 3D case data of SenStick
Sensorized
Chopsticks
Figure 7 Eating activity recognition by sensorized chopsticks
kitchen utensils animals stuffed toy a remote control andany kind of things that can be sensing targets In this sectionwe introduce some case studies results utilizing SenStick
41 Eating Behavior Recognition by Sensorized Chopsticks Inmany cases East Asian people eat meals with chopsticksBy analyzing the data obtained from chopsticks there is apossibility that we can estimate the eating habits includinganswers for questions like ldquohow did you eatrdquo and ldquowhatdid you eatrdquo and so forth These data are very useful formaking healthcare applications and leading a healthy lifeThis is one reason why we develop the sensorized chopsticksusing SenStick shown in Figure 7
Research goals of the eating behavior recognition by sen-sorized chopsticks are as follows G1 (behavior) recognizingthe timing and speed and detecting ldquocroprdquo and ldquointakerdquoaction G2 (food type) classifying food content like ricenoodles and so forth from the eating behavior and G3(person) identifying an individual from using chopsticks Inthis paper we show the research results on G1 and G2
We conducted basic experiments to recognize the eatingbehavior (G1 and G2) We used acceleration and gyroscopedata measured by SenStick where the sensing rate is 10HzThe purpose of experiment 1 is to understand the basic
characteristics of sensing data from sensorized chopstickwhen a participant ate chips The purpose of experiment 2 isto identify the food type from the difference in movement ofchopsticks We assumed 4 classes (C1 rice and dish C2 misosoup C3 noodles and C4 shrimp) as food types
Figure 8 shows time series of sensing data for a set ofeating actions in experiment 1 The eating actions are cropaction and intake action In crop action 6-axis motionsensors detect movement (open and close) of the chopsticksIn intake action sensors detect movement of the arm fromthe dish to the mouth Each graph in Figure 8 shows certaincorrelation with both actions
Figure 9 shows a typical signal for each food in exper-iment 2 In this experiment we calculate the recognitionaccuracy of food type classification by using Random Forestclassier ofWeka Tables 5 and 6 show the classification resultsThe average recognition accuracy was 896 If we did nothave a SenStick it would be hard to develop the sensorizedchopsticks that can be used in the actual environment
42 Daily Activity Monitoring by a Sensorized Belt Recentlycorpulent people are rapidly increasing and about 700 mil-lion people belong to the obesity group [18] Nowadays manyoffice workers use computers heavily and their sitting time
10 Journal of Sensors
Step 1 crop action Step 2 intake action
Lower
Upper
2 4 6 8 10 12 140Elapsed time (sec)
2 4 6 8 10 12 140Elapsed time (sec)
2 4 6 8 10 12 140Elapsed time (sec)
2 4 6 8 10 12 140Elapsed time (sec)
X
Y
Z
Synthetic
X
Y
Z
Synthetic
X
Y
Z
Synthetic
X
Y
Z
Synthetic
minus250
minus150
minus50
50
150
250
Ang
ular
velo
city
(rad
s)
minus20
minus10
01020
Acce
lera
tion
(ms
2)
minus250
minus150
minus50
50
150
250
Ang
ular
velo
city
(rad
s)
minus20
minus10
01020
Acce
lera
tion
(ms
2)
Open1 Close2 Raise3 Eat4 Drop5
1 2 3 4 5
Figure 8 Time series of sensing data in eating activity
in office increases Keeping incorrect postures during longworking time increases the risk of obesity of office workersTherefore it is necessary tomonitor the daily activity of officeworkers for correcting the bad habits
Motivated by this we have developed a sensorized beltcalled Waiston Belt [8] which aims to monitor a userrsquosdaily activities for maintaining health As products of belt-type wearable devices Welt [19] and Belty [20] are on saleThese provide the function of waist measurement activity
tracking and step counting However API for obtainingsensing data is not provided and it is not suitable for researchpurposes Generally it is hard for ordinary researchers toembed sensors battery BLE and vibrator into a small buckle-size case that can be worn in everyday life Nam et al [21]have proposed a belt-attached device for monitoring postureHowever the device has not reached a sufficient level to beused on a daily basis because of exposed wiring and uncoolappearance However by extending SenStick we succeeded
Journal of Sensors 11
C1
C2 C3
C4
Typical signal of each food
C1 rice amp dish C2 miso soup C3 noodles C4 shrimp
400
200
0
minus200
minus400
Figure 9 Eating activity recognition by sensorized chopsticks
SenStick
Waiston Belt
Figure 10 Waiston Belt based on SenStick
Table 5 Experimental result (food type)
Food type Number ofpieces of data Precision Recall 119865-Measure
C1 rice and dish 3011 0894 0945 0919C2 miso soup 2187 0907 0841 0873C3 noodles 2761 088 0896 0888C4 shrimp 2091 0907 088 0893
0896 0896 0895
to develop a stylish prototype of Waiston Belt shown inFigure 10 quickly Since the battery life of SenStick is 15 hourswith a small battery (60mAh) we can sense continuously inour daily lives As shown in Figure 11 it is possible to estimatevarious usersrsquo contexts (How much did you walk Are youusing the stairs Is posture good at work) by analyzing timeseries of sensing data obtained from the sensorized belt
An important point in this use case is that usersrsquo contextscan be estimated by combining multiple sensing data For
Table 6 Confusion matrix of food type recognition
PredictedC1 C2 C3 C4
ActualC1 rice and dish 094 004 001 000C2 miso soup 012 084 004 000C3 noodles 002 002 090 006C4 shrimp 001 001 011 088
example climbing stairs is estimated from the changes ofacceleration and atmospheric pressure The change of thehumidity data will be helpful to detect that a user enteredan air-conditioned room A SenStick makes it possible toeasily carry out research on behavior recognition with acombination of such multiple sensors
We introduce basic experiments about the posture recog-nition during deskwork by using this beltThe purpose of thisexperiment is to identify the posture type from the attitude
12 Journal of Sensors
Enter the lab
HumidityTemp
Humidity dropping
Pressure dropping
Working
Walking WalkingClimbing stairs Sitting (good posture) Sitting (bad posture)
AccsXAccsYAccsZ
60
40
20
1003
1002
1001
1000
999
25
2
15
1
05
0
minus15
minus1
minus05
Air pressure
Figure 11 Time series of sensing data in daily activity
C1 good posture C2 bad posture 1
C3 bad posture 2 C4 bad posture 3
Figure 12 Time series of sensing data in posture monitoring
angle of the belt calculated from 3 axial acceleration sensorsWe set the sensing rate as 10Hz and collected 6000 samples(10min) for each posture We assumed 4 classes (C1 goodposture C2 bad posture 1 C3 bad posture 2 and C4 badposture 3) as shown in Figure 12
In this experiment we calculated the recognition accu-racy of posture type classification by using Random Forestclassifier of Weka Tables 7 and 8 show the classificationresults The average recognition accuracy was 997 Inthe future based on the result of posture estimation we
Journal of Sensors 13
Sensorized glasses
Figure 13 Sensorized glasses
Table 7 Experimental result (posture type)
Posture type Number ofpieces of data Precision Recall 119865-Measure
C1 good posture 6000 1000 1000 1000C2 bad posture 1 6000 0994 0994 0994C3 bad posture 2 6000 0994 0994 0994C4 bad posture 3 6000 0999 0999 0999
0997 0997 0997
will develop a software application that suggests correctinghisher bad posture to a right posture during desk work
43 Indoor Localization by Sensorized Glasses We have pro-posed an indoor localization system using illuminance-basedtrilateration method [22] In this system we assume that theperson wears glasses having an illuminance sensor as shownin Figure 13 and controllable light sources are deployed in thetarget environment Our system controls lights andmeasuresthe variation of illuminance at the glasses As shown inFigure 14 the distances (1198891ndash1198893) from each light (1198971ndash1198973) to theuser (119906) are estimated from the distance-illumination modelconstructed in advance and userrsquos position is estimated basedon trilateration method using estimated distances
As a result of an experiment where three controllablelight sources were deployed in a room we confirmed thatour system could estimate the position within 05m erroron average By using SenStick we could easily develop thesensorized glasses equipped with illuminance sensor
44 Sensorized Airsoft Gun We have developed the sen-sorized airsoft gun for activity recognition in airsoft gameas shown in Figure 15 Airsoft game is a sport in whichparticipants eliminate opponents by hitting each other usingreplica weapons called airsoft guns The graph of a typicalsignal from the sensorized airsoft gun is shown in Figure 16A user can count the number of shots of a ball by attachingSenStick to his airsoft gun
45 Another Example Studentrsquos Unique Achievements andSurvey Results SenStick is also used in a lecture of IoT
Table 8 Confusion matrix of posture type recognition
PredictedC1 C2 C3 C4
ActualC1 good posture 1000 0000 0000 0000C2 bad posture 1 0000 0994 0006 0001C3 bad posture 2 0000 0006 0994 0000C4 bad posture 3 0000 0001 0000 0999
prototyping Students realize their unique IoT devices easilyby using SenSticks Here we introduce some of the developedIoT devices and survey results about the usefulness ofSenStick answered by students
Figure 17 shows a sensorized pencil with two SenSticksThe sensorized pencil is developed to recognize the charac-ters written by this pencil Even though two SenSticks areembedded to detect precise movement of pencil the size ofthe pencil is reasonable
Figure 18 shows a sensorized name tag anddigital signageSince we changed the battery from 60mAh to 900mAh thetag keeps running more than one week It aims to recordemployeesrsquo or studentsrsquo daily activities to support their healthThe digital signage informs the user who wears a sensorizedname tag about the userrsquos activity data when the user standsin front of it
Figure 19 shows a sensorized spoon used in the digitalfish catching game for children In this game the sensorizedspoon detects the scooping movement of the hand Thuschildren can catch the digital fish displayed by the projectionmapping by using sensorized spoonThis systemwas actuallyused at the universityrsquos open campus event and around100 children enjoyed this game and showed interest in IoTtechnology
The survey results obtained from the ten graduate stu-dents (ninemales and one female) who developed prototypesas mentioned above are shown in Table 9 A questionnaireis a five-grade form (5 strongly agree 4 agree a little 3neither agree nor disagree 2 disagree a little and 1 stronglydisagree) From the results of questions 1 and 2 manyparticipants think that SenStick platform is useful for IoT
14 Journal of Sensors
Position
Distance illumination model
Using trilateration method
a
Light 3
Light 2
Light 1
L = minus421d + 1496
L = minus063d + 501
05 1 15 2 25 3 35 4 450Distance between the illuminance sensor and the lighting device (m)
02468
1012141618
Illum
inan
ce v
alue
(lux
)
Distanceu
d1
d2
d3
l1 p1 x1
l3 p3 x3
l2 p2 x2
Figure 14 Indoor localization by sensorized glasses
Sensorized airsoft gun
Figure 15 Sensorized airsoft gun
research and IoT education Also the results of question 3show that many students are interested in the SenStick andwant to use it again From the results of questions 4 and 5we can see that usability and specification of SenStick arerelatively high There were opinions about the usefulness ofSenStick platform such as ldquoI could quickly try out ideas of IoTprototypingrdquo ldquoIt was easy to collect sensing datardquo and ldquoSinceSenStick is small it was easily attached tomost of the thingsrdquoOn the other hand there were opinions about improvements
such as ldquoI want to use higher sensing frequency (50ndash100Hz)rdquoldquoI want to process a sensing data on SenStickrdquo and ldquoI want tomodify the firmware itself depending on the application ofthe IoT devicerdquo
5 Conclusion
In this paper we have introduced a smart and comprehensivesensing platform called SenStick which is composed of
Journal of Sensors 15
Typical signal of airso gun
Hold a gun Ready
Shoot
1 2 3 4 5 6 70
00
30
60
91
21
51
82
12
42
73
03
33
63
94
24
54
85
15
45
76
06
36
66
97
27
57
88
18
48
79
09
39
69
910
210
510
811
111
411
712
012
312
612
913
2
minus600
minus400
minus200
0
200
400
600
800
acc_xacc_yacc_z
gyro_xgyro_y
acc_synthesis gyro_zgyro_synthesis
Figure 16 Time series of sensing data in shooting an airsoft gun
Sensorized pencil
Figure 17 Sensorized pencil
Sensorized name tagVisualization of sensor data
Figure 18 Sensorized name tag
Sensorized spoon Digital fishing game
Figure 19 Sensorized spoon
Table 9 Questionnaire results about SenStick
Question contents ScoreQ1 Do you think that SenStick will be useful for IoTresearch 49
Q2 Do you think that SenStick will be useful for IoTeducation 47
Q3 Do you want to use SenStick again 49Q4 Do you think that SenStick was easy to use 41Q5 Are you satisfied with specification of SenStick 41
hardware software and 3D case data The main purpose ofour platform is to simplify the sensorization of our personalbelongings
The hardware of SenStick is maybe one of the worldrsquossmallest sensor boards which includes 8 kinds of sensorsand BLE In addition to the hardware we also developedsupport software for managing monitoring recording andviewing the hardware By providing 3D case data all theresearchers can measure every tiny activity in daily life easilyand efficiently For an educational use it is also useful forunderstanding sensors and prototyping the desired systemincluding tiny censored things
Conflicts of Interest
The authors declare that there are no conflicts of interestregarding the publication of this paper
Acknowledgments
This work is supported in part by JST PRESTO and JSPSKAKENHI (Grant no 17J10021) and ldquoResearch and Devel-opment of Innovative Network Technologies to Create theFuturerdquo the Commissioned Research of National Instituteof Information and Communications Technology (NICT)Japan
16 Journal of Sensors
References
[1] O Brdiczka M Langet J Maisonnasse and J L CrowleyldquoDetecting human behavior models from multimodal observa-tion in a smart homerdquo IEEE Transactions on Automation Scienceand Engineering vol 6 no 4 pp 588ndash597 2009
[2] J Chen A Kam J Zhang N Liu and L Shue ldquoBathroomactivity monitoring based on soundrdquo in Pervasive ComputingThird International Conference PERVASIVE 2005 MunichGermanyMay 8-13 2005 Proceedings vol 3468 ofLectureNotesin Computer Science pp 47ndash61 Springer Berlin Germany2002
[3] G Bauer K Stockinger and P Lukowicz ldquoRecognizing theUse-Mode of Kitchen Appliances from Their Current Consump-tionrdquo in Smart Sensing and Context vol 5741 of Lecture Notesin Computer Science pp 163ndash176 Springer Berlin HeidelbergBerlin Heidelberg 2009
[4] K Ouchi and M Doi ldquoIndoor-outdoor activity recognition bya smartphonerdquo in Proceedings of the the 2012 ACM Conferencep 537 Pittsburgh Pennsylvania September 2012
[5] A Anjum and M U Ilyas ldquoActivity recognition using smart-phone sensorsrdquo in Proceedings of the IEEE 10th ConsumerCommunications and Networking Conference (CCNC rsquo13) pp914ndash919 IEEE January 2013
[6] O D Lara and M A Labrador ldquoA survey on human activityrecognition using wearable sensorsrdquo IEEE CommunicationsSurveys amp Tutorials vol 15 no 3 pp 1192ndash1209 2013
[7] ldquoJins memerdquo httpsjins-memecom[8] Y Matsuda T Ishioka T Hasegawa et al ldquoWaistonBelt A belt
for monitoring your real abdominal circumference foreverrdquo inProceedings of the ACM International Joint Conference on Perva-sive and Ubiquitous Computing and the 2015 ACM InternationalSymposium on Wearable Computers UbiComp and ISWC 2015pp 329ndash332 September 2015
[9] M Kurahashi K Murao T Terada and M Tsukamoto ldquoPer-sonal identification system based on rotation of toilet paperrollsrdquo in Proceedings of the IEEE International Conference onPervasive Computing and CommunicationsWorkshops (PerComWorkshops rsquo17) pp 521ndash526 IEEE Kona Hawaii USA March2017
[10] ldquoHapiforkrdquo httpswwwhapicom[11] Y Arakawa ldquoSenStick Sensorize every thingsrdquo in Proceedings of
the ACM International Joint Conference on Pervasive and Ubiq-uitous Computing and the 2015 ACM International SymposiumonWearable Computers UbiComp and ISWC 2015 pp 349ndash352September 2015
[12] Y Nakamura T Kanehira Y Arakawa and K YasumotoldquoSenStick 2 Ultra tiny all-in-one sensor with wireless chargingrdquoinProceedings of the 2016ACM International Joint Conference onPervasive and Ubiquitous Computing UbiComp 2016 pp 337ndash340 September 2016
[13] ldquoSensortag cc2541 (texas instruments)rdquo httpwwwticomtoolcc2541dk-sensor
[14] ldquoIot smart modulerdquo httpwwwalpscomjiotsmart[15] ldquoTsnd121 (atr-promotions)rdquo httpwwwatr-pcomproducts
TSND121html[16] ldquoWax9 (axivity)rdquo httpaxivitycomproductwax9[17] Uk biobank ldquoUk biobankrdquo httpwwwukbiobankacuk[18] J C Seidell and J Halberstadt ldquoThe global burden of obesity
and the challenges of preventionrdquo Annals of Nutrition andMetabolism vol 66 supplement 2 pp 7ndash12 2015
[19] ldquoWeltThe smart belt for fashion and healthrdquo httpswwwwelt-corpcom
[20] ldquoBelty power The belt your really deserverdquo httpswwwbeltyco
[21] H Nam J-H Kim and J-I Kim ldquoSmart Belt A wearabledevice for managing abdominal obesityrdquo in Proceedings of theInternational Conference on Big Data and Smart ComputingBigComp 2016 pp 430ndash434 January 2016
[22] K Moriya M Fujimoto Y Arakawa and K Yasumoto ldquoIndoorlocalization based on distance-illuminance model and activecontrol of lighting devicesrdquo in Proceedings of the 2016 Interna-tional Conference on Indoor Positioning and Indoor NavigationIPIN 2016 pp 1ndash6 October 2016
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal of
Volume 201
Submit your manuscripts athttpswwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
6 Journal of Sensors
(32-)
Peripheral
GATT server
SenStick Control Service
Metadata Read Service
Read-write
Data notication
Flashmemory
(0 acceleration 1 gyro 2 magnetic 3 light
4 UV 5 temperaturehumidity 6 pressure)
GATT client
Central
~Sensor Service (0 6)
Figure 3 Generic Attribute (GATT) profile of SenStick
Table 3 BLE communication specification
Type of content SpecAdvertising packet Complete UUID flagScan response Local name (SenStick)Advertising packet interval 100ms (30 s)rarr1000ms (150 s)rarrsleepConnection interval max 80msmin 20msSupervision timeout 4000msSlave latency 0
Basically all the sensing data will be notified to a smart-phone in real time up to a transmission capacity of BLE Ifthe amount of sensing data exceeds the transmission capacitya part of sensing data will be dropped Therefore the real-time sensing data reading function is used formonitoring thestatus or for an interactive application that does not requireprecise data For precise data acquisition the log data readingfunction which guarantees the data reading order is usedThe communication specifications of BLE are described inTable 3
As shown in Table 4 the fixed capacities of a flashmemory are allocated to each sensorThe number of samplesindicates the total number of samples that can be recorded bythe allocated capacities
SenStick stops logging when the number of samples ofone sensor exceeds the storage capacity allocated for it Themaximum recording time of acceleration gyroscope andmagnetic sensor is 4 hours 40 minutes at sampling intervalof 10 milliseconds (sampling frequency 100Hz) and 14 hours10 minutes at sampling period of 30 milliseconds (samplingfrequency 33Hz) The maximum recording time of othersensors is 9 hours 26 minutes at sampling period of 200milliseconds (sampling frequency 5Hz)
32 Software We provide mobile applications for iOS andAndroid and a library for nodejs platform Common func-tions of them are operating function parameter settingfunction and real-time data view function An additional
Table 4 Allocated storage capacity per sensor
Amount of allocatedstorage capacity Samples
Acceleration 102Mbytes 17M samplesGyro 102Mbytes 17M samplesMagnetic 102Mbytes 17M samplesIlluminance 340KB 170K samplesUV (UV) 340KB 170K samplesHumidity 170KB 170K samplesAir temperature 170KB 170K samples
function of iOS application is a firmware update functionThose of Android application and nodejs applications areground truth video recording function and an ability to con-nect multiple SenSticks respectively Since all these softwareapplications are shared on GitHub as an own source softwareall the users can create their application based on their ownpurposes
Communication Protocol Here we explain the communica-tion protocol between SenStick and softwareThe explanationis based on ldquonode-SenStickrdquo which is our developed libraryfor nodejs Figure 4 shows a protocol sequence betweenSenStick and the central terminal (node-SenStick on Mac)
(I) Discovery First of all the application needs to find thesurrounding SenStick by using SenStickdiscovery() methodin our library When SenStickdiscovery() method finds aSenStick a callback function with an instance of SenStickobject containing the SenStickrsquos UUID and LocalName asarguments is called
(II) Connect If the target SenStick is found by the discoveryprocess the application is going to establish a connec-tion with it by calling SenStickprototypeconnectAndSetUp()method After establishing the connection this methodacquires the information necessary for operation from thetarget SenStick
Journal of Sensors 7
Central
Node-SenStick SenStick
SenStickdiscovery()
SenStickprototypeconnectAndSetUp()
Establish connection
SenStickprototypewriteSensorMeasurementCong()
Reect setting
SenStickprototypestartSensingAndLogging()
SenStickprototypenotifySensor()Start sensing
Select sensor to notify
Sensing data
SenStickprototypestopSensingAndLogging()
Stop sensing
Enable log notication
Select log data to notify
SenStickprototypewriteSensorLogReadoutTargetID()
Logging data
SenStickprototypedisconnect()
Disconnect
SenStickprototypenotifySensorLogData()
Periperal
(I) Discovery
(II) Connect
(III) Setting
(V) Obtaining log
(VI) Disconnect
(IV) Start sensingamp
RT receiving
Figure 4 Communication sequence between SenStick and node-SenStick
(III) Setting After establishing the connection the appli-cation sets three parameters (operation mode samplinginterval and measurement range) for each sensor by usingSenStickprototypewriteSensorMeasurementConfig()methodThe operation mode is a parameter that sets valida-tioninvalidation of sensing and logging of each sensor Thesampling interval and measurement range are parameters ofeach sensorwhichwill be changed by the sensing purpose andapplications
(IV) Start Sensing and Real-Time Data Receiving When Sen-StickprototypestartSensingAndLogging() method is calledSenStick starts sensing and logging Only valid sensorsrsquodata are transmitted via BLE and also recorded into aflash memory ID of each log is automatically assignedDuring sensing SenStick notifies sensing data in realtime To receive this notification the application callsSenStickprototypenotifySensor() method (Sensor representseach sensor name) and activates the notification for
8 Journal of Sensors
(i) Frequency(ii) Range
Device select Main view Settings Firmware update
Real-time view Log list in SenStick Download from SenStick
Selected sensorSensor name
Sensing frequency
Range of y-axis
Figure 5 iOS application for SenStick
the sensor When a notification of sensing data isreceived sensorChange event will be issued If Sen-StickprototypestopSensingAndLogging() method is calledSenStick stops sensing and logging
(V) Log Data Reading To read the log data recordedin the flash memory to the application the applicationcalls SenStickprototypenotifySensorLogData()method (Sen-sor represents each sensor name) and enables log datanotification If the application specifies a valid log ID bycalling SenStickprototypewriteSensorLogReadoutTargetID()method SenStick starts informing the sensing data recordedin the log sequentially To receive notification of log datasensorLogDataReceived event will be issued By receiving thisnotification by the application it is possible to acquire the logdata recorded in the memory
(VI) Disconnect After finishing the data acquisition Sen-Stickprototypedisconnect() method will be called to releasethe connection of SenStick And SenStick will begin advertis-ing again
User InterfaceWe explain the user interface of the applicationbased on iOS version Figure 5 shows a sophisticated mobileapplication for SenStick As shown in Figure 5 the userselects target SenStick from devices listThen the applicationshows the main view In the main view the user can selectsensors to be acquired For immediate data monitoring allthe received data via BLE can be shown in the graph view inreal time The application can control all the settings of eachsensor such as sample frequency and sensing range And alsoit has firmware update function Sensing data stored in theflash memory of SenStick is viewed from log list and it can be
downloaded to the smartphone in CSV formatThis functionprevents data loss caused by BLE communication
33 3D Case Data In order to make sensorized things anattachment case for mounting the SenStick is required Wedesigned the basic 3D case of SenStick By extending thisbasic case users can design derivative cases for chopsticksglasses and so on Actually our laboratory has succeeded indesigning cases for several things as shown in Figure 6 Allof these cases data are shared on our community site Eachcase is designed in a few minutes thanks to the basic caseFor example SenStick case for glasses is made by combiningbasic case of SenStick and open 3D data of glasses distributedby i Design Studio In addition due to the popularizationof 3D printers in the world and enhancement of tools andreferences for 3D modeling all the users can print out theirdesired SenStick cases by themselves and attach our sensor toeverything
4 Case Studies of SenStick
SenStick an ultra small sensor can be embedded into varioustiny things in our daily life Our first target is our dailybelongings such as glasses chopsticks and pens becausesensorization of daily belongings is good to intuitively under-standwhat can be sensed andwhich sensor is usefuleffectiveSenStick can upgrade normal glasses to intelligent oneswhichcan monitor a head movement posture and the amount ofsunlight received If SenStick is embedded into chopsticksand pens we can observe eating activities and handwritingactivities respectively In the future we will embed it into astick for elderlymonitoring sports gearsmusical instrument
Journal of Sensors 9
Chopsticks Tooth brush Glasses Umbrella Name card Cookware
A user can design original case data in a few minutes
Basic 3D case of SenStick
Figure 6 3D case data of SenStick
Sensorized
Chopsticks
Figure 7 Eating activity recognition by sensorized chopsticks
kitchen utensils animals stuffed toy a remote control andany kind of things that can be sensing targets In this sectionwe introduce some case studies results utilizing SenStick
41 Eating Behavior Recognition by Sensorized Chopsticks Inmany cases East Asian people eat meals with chopsticksBy analyzing the data obtained from chopsticks there is apossibility that we can estimate the eating habits includinganswers for questions like ldquohow did you eatrdquo and ldquowhatdid you eatrdquo and so forth These data are very useful formaking healthcare applications and leading a healthy lifeThis is one reason why we develop the sensorized chopsticksusing SenStick shown in Figure 7
Research goals of the eating behavior recognition by sen-sorized chopsticks are as follows G1 (behavior) recognizingthe timing and speed and detecting ldquocroprdquo and ldquointakerdquoaction G2 (food type) classifying food content like ricenoodles and so forth from the eating behavior and G3(person) identifying an individual from using chopsticks Inthis paper we show the research results on G1 and G2
We conducted basic experiments to recognize the eatingbehavior (G1 and G2) We used acceleration and gyroscopedata measured by SenStick where the sensing rate is 10HzThe purpose of experiment 1 is to understand the basic
characteristics of sensing data from sensorized chopstickwhen a participant ate chips The purpose of experiment 2 isto identify the food type from the difference in movement ofchopsticks We assumed 4 classes (C1 rice and dish C2 misosoup C3 noodles and C4 shrimp) as food types
Figure 8 shows time series of sensing data for a set ofeating actions in experiment 1 The eating actions are cropaction and intake action In crop action 6-axis motionsensors detect movement (open and close) of the chopsticksIn intake action sensors detect movement of the arm fromthe dish to the mouth Each graph in Figure 8 shows certaincorrelation with both actions
Figure 9 shows a typical signal for each food in exper-iment 2 In this experiment we calculate the recognitionaccuracy of food type classification by using Random Forestclassier ofWeka Tables 5 and 6 show the classification resultsThe average recognition accuracy was 896 If we did nothave a SenStick it would be hard to develop the sensorizedchopsticks that can be used in the actual environment
42 Daily Activity Monitoring by a Sensorized Belt Recentlycorpulent people are rapidly increasing and about 700 mil-lion people belong to the obesity group [18] Nowadays manyoffice workers use computers heavily and their sitting time
10 Journal of Sensors
Step 1 crop action Step 2 intake action
Lower
Upper
2 4 6 8 10 12 140Elapsed time (sec)
2 4 6 8 10 12 140Elapsed time (sec)
2 4 6 8 10 12 140Elapsed time (sec)
2 4 6 8 10 12 140Elapsed time (sec)
X
Y
Z
Synthetic
X
Y
Z
Synthetic
X
Y
Z
Synthetic
X
Y
Z
Synthetic
minus250
minus150
minus50
50
150
250
Ang
ular
velo
city
(rad
s)
minus20
minus10
01020
Acce
lera
tion
(ms
2)
minus250
minus150
minus50
50
150
250
Ang
ular
velo
city
(rad
s)
minus20
minus10
01020
Acce
lera
tion
(ms
2)
Open1 Close2 Raise3 Eat4 Drop5
1 2 3 4 5
Figure 8 Time series of sensing data in eating activity
in office increases Keeping incorrect postures during longworking time increases the risk of obesity of office workersTherefore it is necessary tomonitor the daily activity of officeworkers for correcting the bad habits
Motivated by this we have developed a sensorized beltcalled Waiston Belt [8] which aims to monitor a userrsquosdaily activities for maintaining health As products of belt-type wearable devices Welt [19] and Belty [20] are on saleThese provide the function of waist measurement activity
tracking and step counting However API for obtainingsensing data is not provided and it is not suitable for researchpurposes Generally it is hard for ordinary researchers toembed sensors battery BLE and vibrator into a small buckle-size case that can be worn in everyday life Nam et al [21]have proposed a belt-attached device for monitoring postureHowever the device has not reached a sufficient level to beused on a daily basis because of exposed wiring and uncoolappearance However by extending SenStick we succeeded
Journal of Sensors 11
C1
C2 C3
C4
Typical signal of each food
C1 rice amp dish C2 miso soup C3 noodles C4 shrimp
400
200
0
minus200
minus400
Figure 9 Eating activity recognition by sensorized chopsticks
SenStick
Waiston Belt
Figure 10 Waiston Belt based on SenStick
Table 5 Experimental result (food type)
Food type Number ofpieces of data Precision Recall 119865-Measure
C1 rice and dish 3011 0894 0945 0919C2 miso soup 2187 0907 0841 0873C3 noodles 2761 088 0896 0888C4 shrimp 2091 0907 088 0893
0896 0896 0895
to develop a stylish prototype of Waiston Belt shown inFigure 10 quickly Since the battery life of SenStick is 15 hourswith a small battery (60mAh) we can sense continuously inour daily lives As shown in Figure 11 it is possible to estimatevarious usersrsquo contexts (How much did you walk Are youusing the stairs Is posture good at work) by analyzing timeseries of sensing data obtained from the sensorized belt
An important point in this use case is that usersrsquo contextscan be estimated by combining multiple sensing data For
Table 6 Confusion matrix of food type recognition
PredictedC1 C2 C3 C4
ActualC1 rice and dish 094 004 001 000C2 miso soup 012 084 004 000C3 noodles 002 002 090 006C4 shrimp 001 001 011 088
example climbing stairs is estimated from the changes ofacceleration and atmospheric pressure The change of thehumidity data will be helpful to detect that a user enteredan air-conditioned room A SenStick makes it possible toeasily carry out research on behavior recognition with acombination of such multiple sensors
We introduce basic experiments about the posture recog-nition during deskwork by using this beltThe purpose of thisexperiment is to identify the posture type from the attitude
12 Journal of Sensors
Enter the lab
HumidityTemp
Humidity dropping
Pressure dropping
Working
Walking WalkingClimbing stairs Sitting (good posture) Sitting (bad posture)
AccsXAccsYAccsZ
60
40
20
1003
1002
1001
1000
999
25
2
15
1
05
0
minus15
minus1
minus05
Air pressure
Figure 11 Time series of sensing data in daily activity
C1 good posture C2 bad posture 1
C3 bad posture 2 C4 bad posture 3
Figure 12 Time series of sensing data in posture monitoring
angle of the belt calculated from 3 axial acceleration sensorsWe set the sensing rate as 10Hz and collected 6000 samples(10min) for each posture We assumed 4 classes (C1 goodposture C2 bad posture 1 C3 bad posture 2 and C4 badposture 3) as shown in Figure 12
In this experiment we calculated the recognition accu-racy of posture type classification by using Random Forestclassifier of Weka Tables 7 and 8 show the classificationresults The average recognition accuracy was 997 Inthe future based on the result of posture estimation we
Journal of Sensors 13
Sensorized glasses
Figure 13 Sensorized glasses
Table 7 Experimental result (posture type)
Posture type Number ofpieces of data Precision Recall 119865-Measure
C1 good posture 6000 1000 1000 1000C2 bad posture 1 6000 0994 0994 0994C3 bad posture 2 6000 0994 0994 0994C4 bad posture 3 6000 0999 0999 0999
0997 0997 0997
will develop a software application that suggests correctinghisher bad posture to a right posture during desk work
43 Indoor Localization by Sensorized Glasses We have pro-posed an indoor localization system using illuminance-basedtrilateration method [22] In this system we assume that theperson wears glasses having an illuminance sensor as shownin Figure 13 and controllable light sources are deployed in thetarget environment Our system controls lights andmeasuresthe variation of illuminance at the glasses As shown inFigure 14 the distances (1198891ndash1198893) from each light (1198971ndash1198973) to theuser (119906) are estimated from the distance-illumination modelconstructed in advance and userrsquos position is estimated basedon trilateration method using estimated distances
As a result of an experiment where three controllablelight sources were deployed in a room we confirmed thatour system could estimate the position within 05m erroron average By using SenStick we could easily develop thesensorized glasses equipped with illuminance sensor
44 Sensorized Airsoft Gun We have developed the sen-sorized airsoft gun for activity recognition in airsoft gameas shown in Figure 15 Airsoft game is a sport in whichparticipants eliminate opponents by hitting each other usingreplica weapons called airsoft guns The graph of a typicalsignal from the sensorized airsoft gun is shown in Figure 16A user can count the number of shots of a ball by attachingSenStick to his airsoft gun
45 Another Example Studentrsquos Unique Achievements andSurvey Results SenStick is also used in a lecture of IoT
Table 8 Confusion matrix of posture type recognition
PredictedC1 C2 C3 C4
ActualC1 good posture 1000 0000 0000 0000C2 bad posture 1 0000 0994 0006 0001C3 bad posture 2 0000 0006 0994 0000C4 bad posture 3 0000 0001 0000 0999
prototyping Students realize their unique IoT devices easilyby using SenSticks Here we introduce some of the developedIoT devices and survey results about the usefulness ofSenStick answered by students
Figure 17 shows a sensorized pencil with two SenSticksThe sensorized pencil is developed to recognize the charac-ters written by this pencil Even though two SenSticks areembedded to detect precise movement of pencil the size ofthe pencil is reasonable
Figure 18 shows a sensorized name tag anddigital signageSince we changed the battery from 60mAh to 900mAh thetag keeps running more than one week It aims to recordemployeesrsquo or studentsrsquo daily activities to support their healthThe digital signage informs the user who wears a sensorizedname tag about the userrsquos activity data when the user standsin front of it
Figure 19 shows a sensorized spoon used in the digitalfish catching game for children In this game the sensorizedspoon detects the scooping movement of the hand Thuschildren can catch the digital fish displayed by the projectionmapping by using sensorized spoonThis systemwas actuallyused at the universityrsquos open campus event and around100 children enjoyed this game and showed interest in IoTtechnology
The survey results obtained from the ten graduate stu-dents (ninemales and one female) who developed prototypesas mentioned above are shown in Table 9 A questionnaireis a five-grade form (5 strongly agree 4 agree a little 3neither agree nor disagree 2 disagree a little and 1 stronglydisagree) From the results of questions 1 and 2 manyparticipants think that SenStick platform is useful for IoT
14 Journal of Sensors
Position
Distance illumination model
Using trilateration method
a
Light 3
Light 2
Light 1
L = minus421d + 1496
L = minus063d + 501
05 1 15 2 25 3 35 4 450Distance between the illuminance sensor and the lighting device (m)
02468
1012141618
Illum
inan
ce v
alue
(lux
)
Distanceu
d1
d2
d3
l1 p1 x1
l3 p3 x3
l2 p2 x2
Figure 14 Indoor localization by sensorized glasses
Sensorized airsoft gun
Figure 15 Sensorized airsoft gun
research and IoT education Also the results of question 3show that many students are interested in the SenStick andwant to use it again From the results of questions 4 and 5we can see that usability and specification of SenStick arerelatively high There were opinions about the usefulness ofSenStick platform such as ldquoI could quickly try out ideas of IoTprototypingrdquo ldquoIt was easy to collect sensing datardquo and ldquoSinceSenStick is small it was easily attached tomost of the thingsrdquoOn the other hand there were opinions about improvements
such as ldquoI want to use higher sensing frequency (50ndash100Hz)rdquoldquoI want to process a sensing data on SenStickrdquo and ldquoI want tomodify the firmware itself depending on the application ofthe IoT devicerdquo
5 Conclusion
In this paper we have introduced a smart and comprehensivesensing platform called SenStick which is composed of
Journal of Sensors 15
Typical signal of airso gun
Hold a gun Ready
Shoot
1 2 3 4 5 6 70
00
30
60
91
21
51
82
12
42
73
03
33
63
94
24
54
85
15
45
76
06
36
66
97
27
57
88
18
48
79
09
39
69
910
210
510
811
111
411
712
012
312
612
913
2
minus600
minus400
minus200
0
200
400
600
800
acc_xacc_yacc_z
gyro_xgyro_y
acc_synthesis gyro_zgyro_synthesis
Figure 16 Time series of sensing data in shooting an airsoft gun
Sensorized pencil
Figure 17 Sensorized pencil
Sensorized name tagVisualization of sensor data
Figure 18 Sensorized name tag
Sensorized spoon Digital fishing game
Figure 19 Sensorized spoon
Table 9 Questionnaire results about SenStick
Question contents ScoreQ1 Do you think that SenStick will be useful for IoTresearch 49
Q2 Do you think that SenStick will be useful for IoTeducation 47
Q3 Do you want to use SenStick again 49Q4 Do you think that SenStick was easy to use 41Q5 Are you satisfied with specification of SenStick 41
hardware software and 3D case data The main purpose ofour platform is to simplify the sensorization of our personalbelongings
The hardware of SenStick is maybe one of the worldrsquossmallest sensor boards which includes 8 kinds of sensorsand BLE In addition to the hardware we also developedsupport software for managing monitoring recording andviewing the hardware By providing 3D case data all theresearchers can measure every tiny activity in daily life easilyand efficiently For an educational use it is also useful forunderstanding sensors and prototyping the desired systemincluding tiny censored things
Conflicts of Interest
The authors declare that there are no conflicts of interestregarding the publication of this paper
Acknowledgments
This work is supported in part by JST PRESTO and JSPSKAKENHI (Grant no 17J10021) and ldquoResearch and Devel-opment of Innovative Network Technologies to Create theFuturerdquo the Commissioned Research of National Instituteof Information and Communications Technology (NICT)Japan
16 Journal of Sensors
References
[1] O Brdiczka M Langet J Maisonnasse and J L CrowleyldquoDetecting human behavior models from multimodal observa-tion in a smart homerdquo IEEE Transactions on Automation Scienceand Engineering vol 6 no 4 pp 588ndash597 2009
[2] J Chen A Kam J Zhang N Liu and L Shue ldquoBathroomactivity monitoring based on soundrdquo in Pervasive ComputingThird International Conference PERVASIVE 2005 MunichGermanyMay 8-13 2005 Proceedings vol 3468 ofLectureNotesin Computer Science pp 47ndash61 Springer Berlin Germany2002
[3] G Bauer K Stockinger and P Lukowicz ldquoRecognizing theUse-Mode of Kitchen Appliances from Their Current Consump-tionrdquo in Smart Sensing and Context vol 5741 of Lecture Notesin Computer Science pp 163ndash176 Springer Berlin HeidelbergBerlin Heidelberg 2009
[4] K Ouchi and M Doi ldquoIndoor-outdoor activity recognition bya smartphonerdquo in Proceedings of the the 2012 ACM Conferencep 537 Pittsburgh Pennsylvania September 2012
[5] A Anjum and M U Ilyas ldquoActivity recognition using smart-phone sensorsrdquo in Proceedings of the IEEE 10th ConsumerCommunications and Networking Conference (CCNC rsquo13) pp914ndash919 IEEE January 2013
[6] O D Lara and M A Labrador ldquoA survey on human activityrecognition using wearable sensorsrdquo IEEE CommunicationsSurveys amp Tutorials vol 15 no 3 pp 1192ndash1209 2013
[7] ldquoJins memerdquo httpsjins-memecom[8] Y Matsuda T Ishioka T Hasegawa et al ldquoWaistonBelt A belt
for monitoring your real abdominal circumference foreverrdquo inProceedings of the ACM International Joint Conference on Perva-sive and Ubiquitous Computing and the 2015 ACM InternationalSymposium on Wearable Computers UbiComp and ISWC 2015pp 329ndash332 September 2015
[9] M Kurahashi K Murao T Terada and M Tsukamoto ldquoPer-sonal identification system based on rotation of toilet paperrollsrdquo in Proceedings of the IEEE International Conference onPervasive Computing and CommunicationsWorkshops (PerComWorkshops rsquo17) pp 521ndash526 IEEE Kona Hawaii USA March2017
[10] ldquoHapiforkrdquo httpswwwhapicom[11] Y Arakawa ldquoSenStick Sensorize every thingsrdquo in Proceedings of
the ACM International Joint Conference on Pervasive and Ubiq-uitous Computing and the 2015 ACM International SymposiumonWearable Computers UbiComp and ISWC 2015 pp 349ndash352September 2015
[12] Y Nakamura T Kanehira Y Arakawa and K YasumotoldquoSenStick 2 Ultra tiny all-in-one sensor with wireless chargingrdquoinProceedings of the 2016ACM International Joint Conference onPervasive and Ubiquitous Computing UbiComp 2016 pp 337ndash340 September 2016
[13] ldquoSensortag cc2541 (texas instruments)rdquo httpwwwticomtoolcc2541dk-sensor
[14] ldquoIot smart modulerdquo httpwwwalpscomjiotsmart[15] ldquoTsnd121 (atr-promotions)rdquo httpwwwatr-pcomproducts
TSND121html[16] ldquoWax9 (axivity)rdquo httpaxivitycomproductwax9[17] Uk biobank ldquoUk biobankrdquo httpwwwukbiobankacuk[18] J C Seidell and J Halberstadt ldquoThe global burden of obesity
and the challenges of preventionrdquo Annals of Nutrition andMetabolism vol 66 supplement 2 pp 7ndash12 2015
[19] ldquoWeltThe smart belt for fashion and healthrdquo httpswwwwelt-corpcom
[20] ldquoBelty power The belt your really deserverdquo httpswwwbeltyco
[21] H Nam J-H Kim and J-I Kim ldquoSmart Belt A wearabledevice for managing abdominal obesityrdquo in Proceedings of theInternational Conference on Big Data and Smart ComputingBigComp 2016 pp 430ndash434 January 2016
[22] K Moriya M Fujimoto Y Arakawa and K Yasumoto ldquoIndoorlocalization based on distance-illuminance model and activecontrol of lighting devicesrdquo in Proceedings of the 2016 Interna-tional Conference on Indoor Positioning and Indoor NavigationIPIN 2016 pp 1ndash6 October 2016
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal of
Volume 201
Submit your manuscripts athttpswwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
Journal of Sensors 7
Central
Node-SenStick SenStick
SenStickdiscovery()
SenStickprototypeconnectAndSetUp()
Establish connection
SenStickprototypewriteSensorMeasurementCong()
Reect setting
SenStickprototypestartSensingAndLogging()
SenStickprototypenotifySensor()Start sensing
Select sensor to notify
Sensing data
SenStickprototypestopSensingAndLogging()
Stop sensing
Enable log notication
Select log data to notify
SenStickprototypewriteSensorLogReadoutTargetID()
Logging data
SenStickprototypedisconnect()
Disconnect
SenStickprototypenotifySensorLogData()
Periperal
(I) Discovery
(II) Connect
(III) Setting
(V) Obtaining log
(VI) Disconnect
(IV) Start sensingamp
RT receiving
Figure 4 Communication sequence between SenStick and node-SenStick
(III) Setting After establishing the connection the appli-cation sets three parameters (operation mode samplinginterval and measurement range) for each sensor by usingSenStickprototypewriteSensorMeasurementConfig()methodThe operation mode is a parameter that sets valida-tioninvalidation of sensing and logging of each sensor Thesampling interval and measurement range are parameters ofeach sensorwhichwill be changed by the sensing purpose andapplications
(IV) Start Sensing and Real-Time Data Receiving When Sen-StickprototypestartSensingAndLogging() method is calledSenStick starts sensing and logging Only valid sensorsrsquodata are transmitted via BLE and also recorded into aflash memory ID of each log is automatically assignedDuring sensing SenStick notifies sensing data in realtime To receive this notification the application callsSenStickprototypenotifySensor() method (Sensor representseach sensor name) and activates the notification for
8 Journal of Sensors
(i) Frequency(ii) Range
Device select Main view Settings Firmware update
Real-time view Log list in SenStick Download from SenStick
Selected sensorSensor name
Sensing frequency
Range of y-axis
Figure 5 iOS application for SenStick
the sensor When a notification of sensing data isreceived sensorChange event will be issued If Sen-StickprototypestopSensingAndLogging() method is calledSenStick stops sensing and logging
(V) Log Data Reading To read the log data recordedin the flash memory to the application the applicationcalls SenStickprototypenotifySensorLogData()method (Sen-sor represents each sensor name) and enables log datanotification If the application specifies a valid log ID bycalling SenStickprototypewriteSensorLogReadoutTargetID()method SenStick starts informing the sensing data recordedin the log sequentially To receive notification of log datasensorLogDataReceived event will be issued By receiving thisnotification by the application it is possible to acquire the logdata recorded in the memory
(VI) Disconnect After finishing the data acquisition Sen-Stickprototypedisconnect() method will be called to releasethe connection of SenStick And SenStick will begin advertis-ing again
User InterfaceWe explain the user interface of the applicationbased on iOS version Figure 5 shows a sophisticated mobileapplication for SenStick As shown in Figure 5 the userselects target SenStick from devices listThen the applicationshows the main view In the main view the user can selectsensors to be acquired For immediate data monitoring allthe received data via BLE can be shown in the graph view inreal time The application can control all the settings of eachsensor such as sample frequency and sensing range And alsoit has firmware update function Sensing data stored in theflash memory of SenStick is viewed from log list and it can be
downloaded to the smartphone in CSV formatThis functionprevents data loss caused by BLE communication
33 3D Case Data In order to make sensorized things anattachment case for mounting the SenStick is required Wedesigned the basic 3D case of SenStick By extending thisbasic case users can design derivative cases for chopsticksglasses and so on Actually our laboratory has succeeded indesigning cases for several things as shown in Figure 6 Allof these cases data are shared on our community site Eachcase is designed in a few minutes thanks to the basic caseFor example SenStick case for glasses is made by combiningbasic case of SenStick and open 3D data of glasses distributedby i Design Studio In addition due to the popularizationof 3D printers in the world and enhancement of tools andreferences for 3D modeling all the users can print out theirdesired SenStick cases by themselves and attach our sensor toeverything
4 Case Studies of SenStick
SenStick an ultra small sensor can be embedded into varioustiny things in our daily life Our first target is our dailybelongings such as glasses chopsticks and pens becausesensorization of daily belongings is good to intuitively under-standwhat can be sensed andwhich sensor is usefuleffectiveSenStick can upgrade normal glasses to intelligent oneswhichcan monitor a head movement posture and the amount ofsunlight received If SenStick is embedded into chopsticksand pens we can observe eating activities and handwritingactivities respectively In the future we will embed it into astick for elderlymonitoring sports gearsmusical instrument
Journal of Sensors 9
Chopsticks Tooth brush Glasses Umbrella Name card Cookware
A user can design original case data in a few minutes
Basic 3D case of SenStick
Figure 6 3D case data of SenStick
Sensorized
Chopsticks
Figure 7 Eating activity recognition by sensorized chopsticks
kitchen utensils animals stuffed toy a remote control andany kind of things that can be sensing targets In this sectionwe introduce some case studies results utilizing SenStick
41 Eating Behavior Recognition by Sensorized Chopsticks Inmany cases East Asian people eat meals with chopsticksBy analyzing the data obtained from chopsticks there is apossibility that we can estimate the eating habits includinganswers for questions like ldquohow did you eatrdquo and ldquowhatdid you eatrdquo and so forth These data are very useful formaking healthcare applications and leading a healthy lifeThis is one reason why we develop the sensorized chopsticksusing SenStick shown in Figure 7
Research goals of the eating behavior recognition by sen-sorized chopsticks are as follows G1 (behavior) recognizingthe timing and speed and detecting ldquocroprdquo and ldquointakerdquoaction G2 (food type) classifying food content like ricenoodles and so forth from the eating behavior and G3(person) identifying an individual from using chopsticks Inthis paper we show the research results on G1 and G2
We conducted basic experiments to recognize the eatingbehavior (G1 and G2) We used acceleration and gyroscopedata measured by SenStick where the sensing rate is 10HzThe purpose of experiment 1 is to understand the basic
characteristics of sensing data from sensorized chopstickwhen a participant ate chips The purpose of experiment 2 isto identify the food type from the difference in movement ofchopsticks We assumed 4 classes (C1 rice and dish C2 misosoup C3 noodles and C4 shrimp) as food types
Figure 8 shows time series of sensing data for a set ofeating actions in experiment 1 The eating actions are cropaction and intake action In crop action 6-axis motionsensors detect movement (open and close) of the chopsticksIn intake action sensors detect movement of the arm fromthe dish to the mouth Each graph in Figure 8 shows certaincorrelation with both actions
Figure 9 shows a typical signal for each food in exper-iment 2 In this experiment we calculate the recognitionaccuracy of food type classification by using Random Forestclassier ofWeka Tables 5 and 6 show the classification resultsThe average recognition accuracy was 896 If we did nothave a SenStick it would be hard to develop the sensorizedchopsticks that can be used in the actual environment
42 Daily Activity Monitoring by a Sensorized Belt Recentlycorpulent people are rapidly increasing and about 700 mil-lion people belong to the obesity group [18] Nowadays manyoffice workers use computers heavily and their sitting time
10 Journal of Sensors
Step 1 crop action Step 2 intake action
Lower
Upper
2 4 6 8 10 12 140Elapsed time (sec)
2 4 6 8 10 12 140Elapsed time (sec)
2 4 6 8 10 12 140Elapsed time (sec)
2 4 6 8 10 12 140Elapsed time (sec)
X
Y
Z
Synthetic
X
Y
Z
Synthetic
X
Y
Z
Synthetic
X
Y
Z
Synthetic
minus250
minus150
minus50
50
150
250
Ang
ular
velo
city
(rad
s)
minus20
minus10
01020
Acce
lera
tion
(ms
2)
minus250
minus150
minus50
50
150
250
Ang
ular
velo
city
(rad
s)
minus20
minus10
01020
Acce
lera
tion
(ms
2)
Open1 Close2 Raise3 Eat4 Drop5
1 2 3 4 5
Figure 8 Time series of sensing data in eating activity
in office increases Keeping incorrect postures during longworking time increases the risk of obesity of office workersTherefore it is necessary tomonitor the daily activity of officeworkers for correcting the bad habits
Motivated by this we have developed a sensorized beltcalled Waiston Belt [8] which aims to monitor a userrsquosdaily activities for maintaining health As products of belt-type wearable devices Welt [19] and Belty [20] are on saleThese provide the function of waist measurement activity
tracking and step counting However API for obtainingsensing data is not provided and it is not suitable for researchpurposes Generally it is hard for ordinary researchers toembed sensors battery BLE and vibrator into a small buckle-size case that can be worn in everyday life Nam et al [21]have proposed a belt-attached device for monitoring postureHowever the device has not reached a sufficient level to beused on a daily basis because of exposed wiring and uncoolappearance However by extending SenStick we succeeded
Journal of Sensors 11
C1
C2 C3
C4
Typical signal of each food
C1 rice amp dish C2 miso soup C3 noodles C4 shrimp
400
200
0
minus200
minus400
Figure 9 Eating activity recognition by sensorized chopsticks
SenStick
Waiston Belt
Figure 10 Waiston Belt based on SenStick
Table 5 Experimental result (food type)
Food type Number ofpieces of data Precision Recall 119865-Measure
C1 rice and dish 3011 0894 0945 0919C2 miso soup 2187 0907 0841 0873C3 noodles 2761 088 0896 0888C4 shrimp 2091 0907 088 0893
0896 0896 0895
to develop a stylish prototype of Waiston Belt shown inFigure 10 quickly Since the battery life of SenStick is 15 hourswith a small battery (60mAh) we can sense continuously inour daily lives As shown in Figure 11 it is possible to estimatevarious usersrsquo contexts (How much did you walk Are youusing the stairs Is posture good at work) by analyzing timeseries of sensing data obtained from the sensorized belt
An important point in this use case is that usersrsquo contextscan be estimated by combining multiple sensing data For
Table 6 Confusion matrix of food type recognition
PredictedC1 C2 C3 C4
ActualC1 rice and dish 094 004 001 000C2 miso soup 012 084 004 000C3 noodles 002 002 090 006C4 shrimp 001 001 011 088
example climbing stairs is estimated from the changes ofacceleration and atmospheric pressure The change of thehumidity data will be helpful to detect that a user enteredan air-conditioned room A SenStick makes it possible toeasily carry out research on behavior recognition with acombination of such multiple sensors
We introduce basic experiments about the posture recog-nition during deskwork by using this beltThe purpose of thisexperiment is to identify the posture type from the attitude
12 Journal of Sensors
Enter the lab
HumidityTemp
Humidity dropping
Pressure dropping
Working
Walking WalkingClimbing stairs Sitting (good posture) Sitting (bad posture)
AccsXAccsYAccsZ
60
40
20
1003
1002
1001
1000
999
25
2
15
1
05
0
minus15
minus1
minus05
Air pressure
Figure 11 Time series of sensing data in daily activity
C1 good posture C2 bad posture 1
C3 bad posture 2 C4 bad posture 3
Figure 12 Time series of sensing data in posture monitoring
angle of the belt calculated from 3 axial acceleration sensorsWe set the sensing rate as 10Hz and collected 6000 samples(10min) for each posture We assumed 4 classes (C1 goodposture C2 bad posture 1 C3 bad posture 2 and C4 badposture 3) as shown in Figure 12
In this experiment we calculated the recognition accu-racy of posture type classification by using Random Forestclassifier of Weka Tables 7 and 8 show the classificationresults The average recognition accuracy was 997 Inthe future based on the result of posture estimation we
Journal of Sensors 13
Sensorized glasses
Figure 13 Sensorized glasses
Table 7 Experimental result (posture type)
Posture type Number ofpieces of data Precision Recall 119865-Measure
C1 good posture 6000 1000 1000 1000C2 bad posture 1 6000 0994 0994 0994C3 bad posture 2 6000 0994 0994 0994C4 bad posture 3 6000 0999 0999 0999
0997 0997 0997
will develop a software application that suggests correctinghisher bad posture to a right posture during desk work
43 Indoor Localization by Sensorized Glasses We have pro-posed an indoor localization system using illuminance-basedtrilateration method [22] In this system we assume that theperson wears glasses having an illuminance sensor as shownin Figure 13 and controllable light sources are deployed in thetarget environment Our system controls lights andmeasuresthe variation of illuminance at the glasses As shown inFigure 14 the distances (1198891ndash1198893) from each light (1198971ndash1198973) to theuser (119906) are estimated from the distance-illumination modelconstructed in advance and userrsquos position is estimated basedon trilateration method using estimated distances
As a result of an experiment where three controllablelight sources were deployed in a room we confirmed thatour system could estimate the position within 05m erroron average By using SenStick we could easily develop thesensorized glasses equipped with illuminance sensor
44 Sensorized Airsoft Gun We have developed the sen-sorized airsoft gun for activity recognition in airsoft gameas shown in Figure 15 Airsoft game is a sport in whichparticipants eliminate opponents by hitting each other usingreplica weapons called airsoft guns The graph of a typicalsignal from the sensorized airsoft gun is shown in Figure 16A user can count the number of shots of a ball by attachingSenStick to his airsoft gun
45 Another Example Studentrsquos Unique Achievements andSurvey Results SenStick is also used in a lecture of IoT
Table 8 Confusion matrix of posture type recognition
PredictedC1 C2 C3 C4
ActualC1 good posture 1000 0000 0000 0000C2 bad posture 1 0000 0994 0006 0001C3 bad posture 2 0000 0006 0994 0000C4 bad posture 3 0000 0001 0000 0999
prototyping Students realize their unique IoT devices easilyby using SenSticks Here we introduce some of the developedIoT devices and survey results about the usefulness ofSenStick answered by students
Figure 17 shows a sensorized pencil with two SenSticksThe sensorized pencil is developed to recognize the charac-ters written by this pencil Even though two SenSticks areembedded to detect precise movement of pencil the size ofthe pencil is reasonable
Figure 18 shows a sensorized name tag anddigital signageSince we changed the battery from 60mAh to 900mAh thetag keeps running more than one week It aims to recordemployeesrsquo or studentsrsquo daily activities to support their healthThe digital signage informs the user who wears a sensorizedname tag about the userrsquos activity data when the user standsin front of it
Figure 19 shows a sensorized spoon used in the digitalfish catching game for children In this game the sensorizedspoon detects the scooping movement of the hand Thuschildren can catch the digital fish displayed by the projectionmapping by using sensorized spoonThis systemwas actuallyused at the universityrsquos open campus event and around100 children enjoyed this game and showed interest in IoTtechnology
The survey results obtained from the ten graduate stu-dents (ninemales and one female) who developed prototypesas mentioned above are shown in Table 9 A questionnaireis a five-grade form (5 strongly agree 4 agree a little 3neither agree nor disagree 2 disagree a little and 1 stronglydisagree) From the results of questions 1 and 2 manyparticipants think that SenStick platform is useful for IoT
14 Journal of Sensors
Position
Distance illumination model
Using trilateration method
a
Light 3
Light 2
Light 1
L = minus421d + 1496
L = minus063d + 501
05 1 15 2 25 3 35 4 450Distance between the illuminance sensor and the lighting device (m)
02468
1012141618
Illum
inan
ce v
alue
(lux
)
Distanceu
d1
d2
d3
l1 p1 x1
l3 p3 x3
l2 p2 x2
Figure 14 Indoor localization by sensorized glasses
Sensorized airsoft gun
Figure 15 Sensorized airsoft gun
research and IoT education Also the results of question 3show that many students are interested in the SenStick andwant to use it again From the results of questions 4 and 5we can see that usability and specification of SenStick arerelatively high There were opinions about the usefulness ofSenStick platform such as ldquoI could quickly try out ideas of IoTprototypingrdquo ldquoIt was easy to collect sensing datardquo and ldquoSinceSenStick is small it was easily attached tomost of the thingsrdquoOn the other hand there were opinions about improvements
such as ldquoI want to use higher sensing frequency (50ndash100Hz)rdquoldquoI want to process a sensing data on SenStickrdquo and ldquoI want tomodify the firmware itself depending on the application ofthe IoT devicerdquo
5 Conclusion
In this paper we have introduced a smart and comprehensivesensing platform called SenStick which is composed of
Journal of Sensors 15
Typical signal of airso gun
Hold a gun Ready
Shoot
1 2 3 4 5 6 70
00
30
60
91
21
51
82
12
42
73
03
33
63
94
24
54
85
15
45
76
06
36
66
97
27
57
88
18
48
79
09
39
69
910
210
510
811
111
411
712
012
312
612
913
2
minus600
minus400
minus200
0
200
400
600
800
acc_xacc_yacc_z
gyro_xgyro_y
acc_synthesis gyro_zgyro_synthesis
Figure 16 Time series of sensing data in shooting an airsoft gun
Sensorized pencil
Figure 17 Sensorized pencil
Sensorized name tagVisualization of sensor data
Figure 18 Sensorized name tag
Sensorized spoon Digital fishing game
Figure 19 Sensorized spoon
Table 9 Questionnaire results about SenStick
Question contents ScoreQ1 Do you think that SenStick will be useful for IoTresearch 49
Q2 Do you think that SenStick will be useful for IoTeducation 47
Q3 Do you want to use SenStick again 49Q4 Do you think that SenStick was easy to use 41Q5 Are you satisfied with specification of SenStick 41
hardware software and 3D case data The main purpose ofour platform is to simplify the sensorization of our personalbelongings
The hardware of SenStick is maybe one of the worldrsquossmallest sensor boards which includes 8 kinds of sensorsand BLE In addition to the hardware we also developedsupport software for managing monitoring recording andviewing the hardware By providing 3D case data all theresearchers can measure every tiny activity in daily life easilyand efficiently For an educational use it is also useful forunderstanding sensors and prototyping the desired systemincluding tiny censored things
Conflicts of Interest
The authors declare that there are no conflicts of interestregarding the publication of this paper
Acknowledgments
This work is supported in part by JST PRESTO and JSPSKAKENHI (Grant no 17J10021) and ldquoResearch and Devel-opment of Innovative Network Technologies to Create theFuturerdquo the Commissioned Research of National Instituteof Information and Communications Technology (NICT)Japan
16 Journal of Sensors
References
[1] O Brdiczka M Langet J Maisonnasse and J L CrowleyldquoDetecting human behavior models from multimodal observa-tion in a smart homerdquo IEEE Transactions on Automation Scienceand Engineering vol 6 no 4 pp 588ndash597 2009
[2] J Chen A Kam J Zhang N Liu and L Shue ldquoBathroomactivity monitoring based on soundrdquo in Pervasive ComputingThird International Conference PERVASIVE 2005 MunichGermanyMay 8-13 2005 Proceedings vol 3468 ofLectureNotesin Computer Science pp 47ndash61 Springer Berlin Germany2002
[3] G Bauer K Stockinger and P Lukowicz ldquoRecognizing theUse-Mode of Kitchen Appliances from Their Current Consump-tionrdquo in Smart Sensing and Context vol 5741 of Lecture Notesin Computer Science pp 163ndash176 Springer Berlin HeidelbergBerlin Heidelberg 2009
[4] K Ouchi and M Doi ldquoIndoor-outdoor activity recognition bya smartphonerdquo in Proceedings of the the 2012 ACM Conferencep 537 Pittsburgh Pennsylvania September 2012
[5] A Anjum and M U Ilyas ldquoActivity recognition using smart-phone sensorsrdquo in Proceedings of the IEEE 10th ConsumerCommunications and Networking Conference (CCNC rsquo13) pp914ndash919 IEEE January 2013
[6] O D Lara and M A Labrador ldquoA survey on human activityrecognition using wearable sensorsrdquo IEEE CommunicationsSurveys amp Tutorials vol 15 no 3 pp 1192ndash1209 2013
[7] ldquoJins memerdquo httpsjins-memecom[8] Y Matsuda T Ishioka T Hasegawa et al ldquoWaistonBelt A belt
for monitoring your real abdominal circumference foreverrdquo inProceedings of the ACM International Joint Conference on Perva-sive and Ubiquitous Computing and the 2015 ACM InternationalSymposium on Wearable Computers UbiComp and ISWC 2015pp 329ndash332 September 2015
[9] M Kurahashi K Murao T Terada and M Tsukamoto ldquoPer-sonal identification system based on rotation of toilet paperrollsrdquo in Proceedings of the IEEE International Conference onPervasive Computing and CommunicationsWorkshops (PerComWorkshops rsquo17) pp 521ndash526 IEEE Kona Hawaii USA March2017
[10] ldquoHapiforkrdquo httpswwwhapicom[11] Y Arakawa ldquoSenStick Sensorize every thingsrdquo in Proceedings of
the ACM International Joint Conference on Pervasive and Ubiq-uitous Computing and the 2015 ACM International SymposiumonWearable Computers UbiComp and ISWC 2015 pp 349ndash352September 2015
[12] Y Nakamura T Kanehira Y Arakawa and K YasumotoldquoSenStick 2 Ultra tiny all-in-one sensor with wireless chargingrdquoinProceedings of the 2016ACM International Joint Conference onPervasive and Ubiquitous Computing UbiComp 2016 pp 337ndash340 September 2016
[13] ldquoSensortag cc2541 (texas instruments)rdquo httpwwwticomtoolcc2541dk-sensor
[14] ldquoIot smart modulerdquo httpwwwalpscomjiotsmart[15] ldquoTsnd121 (atr-promotions)rdquo httpwwwatr-pcomproducts
TSND121html[16] ldquoWax9 (axivity)rdquo httpaxivitycomproductwax9[17] Uk biobank ldquoUk biobankrdquo httpwwwukbiobankacuk[18] J C Seidell and J Halberstadt ldquoThe global burden of obesity
and the challenges of preventionrdquo Annals of Nutrition andMetabolism vol 66 supplement 2 pp 7ndash12 2015
[19] ldquoWeltThe smart belt for fashion and healthrdquo httpswwwwelt-corpcom
[20] ldquoBelty power The belt your really deserverdquo httpswwwbeltyco
[21] H Nam J-H Kim and J-I Kim ldquoSmart Belt A wearabledevice for managing abdominal obesityrdquo in Proceedings of theInternational Conference on Big Data and Smart ComputingBigComp 2016 pp 430ndash434 January 2016
[22] K Moriya M Fujimoto Y Arakawa and K Yasumoto ldquoIndoorlocalization based on distance-illuminance model and activecontrol of lighting devicesrdquo in Proceedings of the 2016 Interna-tional Conference on Indoor Positioning and Indoor NavigationIPIN 2016 pp 1ndash6 October 2016
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal of
Volume 201
Submit your manuscripts athttpswwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
8 Journal of Sensors
(i) Frequency(ii) Range
Device select Main view Settings Firmware update
Real-time view Log list in SenStick Download from SenStick
Selected sensorSensor name
Sensing frequency
Range of y-axis
Figure 5 iOS application for SenStick
the sensor When a notification of sensing data isreceived sensorChange event will be issued If Sen-StickprototypestopSensingAndLogging() method is calledSenStick stops sensing and logging
(V) Log Data Reading To read the log data recordedin the flash memory to the application the applicationcalls SenStickprototypenotifySensorLogData()method (Sen-sor represents each sensor name) and enables log datanotification If the application specifies a valid log ID bycalling SenStickprototypewriteSensorLogReadoutTargetID()method SenStick starts informing the sensing data recordedin the log sequentially To receive notification of log datasensorLogDataReceived event will be issued By receiving thisnotification by the application it is possible to acquire the logdata recorded in the memory
(VI) Disconnect After finishing the data acquisition Sen-Stickprototypedisconnect() method will be called to releasethe connection of SenStick And SenStick will begin advertis-ing again
User InterfaceWe explain the user interface of the applicationbased on iOS version Figure 5 shows a sophisticated mobileapplication for SenStick As shown in Figure 5 the userselects target SenStick from devices listThen the applicationshows the main view In the main view the user can selectsensors to be acquired For immediate data monitoring allthe received data via BLE can be shown in the graph view inreal time The application can control all the settings of eachsensor such as sample frequency and sensing range And alsoit has firmware update function Sensing data stored in theflash memory of SenStick is viewed from log list and it can be
downloaded to the smartphone in CSV formatThis functionprevents data loss caused by BLE communication
33 3D Case Data In order to make sensorized things anattachment case for mounting the SenStick is required Wedesigned the basic 3D case of SenStick By extending thisbasic case users can design derivative cases for chopsticksglasses and so on Actually our laboratory has succeeded indesigning cases for several things as shown in Figure 6 Allof these cases data are shared on our community site Eachcase is designed in a few minutes thanks to the basic caseFor example SenStick case for glasses is made by combiningbasic case of SenStick and open 3D data of glasses distributedby i Design Studio In addition due to the popularizationof 3D printers in the world and enhancement of tools andreferences for 3D modeling all the users can print out theirdesired SenStick cases by themselves and attach our sensor toeverything
4 Case Studies of SenStick
SenStick an ultra small sensor can be embedded into varioustiny things in our daily life Our first target is our dailybelongings such as glasses chopsticks and pens becausesensorization of daily belongings is good to intuitively under-standwhat can be sensed andwhich sensor is usefuleffectiveSenStick can upgrade normal glasses to intelligent oneswhichcan monitor a head movement posture and the amount ofsunlight received If SenStick is embedded into chopsticksand pens we can observe eating activities and handwritingactivities respectively In the future we will embed it into astick for elderlymonitoring sports gearsmusical instrument
Journal of Sensors 9
Chopsticks Tooth brush Glasses Umbrella Name card Cookware
A user can design original case data in a few minutes
Basic 3D case of SenStick
Figure 6 3D case data of SenStick
Sensorized
Chopsticks
Figure 7 Eating activity recognition by sensorized chopsticks
kitchen utensils animals stuffed toy a remote control andany kind of things that can be sensing targets In this sectionwe introduce some case studies results utilizing SenStick
41 Eating Behavior Recognition by Sensorized Chopsticks Inmany cases East Asian people eat meals with chopsticksBy analyzing the data obtained from chopsticks there is apossibility that we can estimate the eating habits includinganswers for questions like ldquohow did you eatrdquo and ldquowhatdid you eatrdquo and so forth These data are very useful formaking healthcare applications and leading a healthy lifeThis is one reason why we develop the sensorized chopsticksusing SenStick shown in Figure 7
Research goals of the eating behavior recognition by sen-sorized chopsticks are as follows G1 (behavior) recognizingthe timing and speed and detecting ldquocroprdquo and ldquointakerdquoaction G2 (food type) classifying food content like ricenoodles and so forth from the eating behavior and G3(person) identifying an individual from using chopsticks Inthis paper we show the research results on G1 and G2
We conducted basic experiments to recognize the eatingbehavior (G1 and G2) We used acceleration and gyroscopedata measured by SenStick where the sensing rate is 10HzThe purpose of experiment 1 is to understand the basic
characteristics of sensing data from sensorized chopstickwhen a participant ate chips The purpose of experiment 2 isto identify the food type from the difference in movement ofchopsticks We assumed 4 classes (C1 rice and dish C2 misosoup C3 noodles and C4 shrimp) as food types
Figure 8 shows time series of sensing data for a set ofeating actions in experiment 1 The eating actions are cropaction and intake action In crop action 6-axis motionsensors detect movement (open and close) of the chopsticksIn intake action sensors detect movement of the arm fromthe dish to the mouth Each graph in Figure 8 shows certaincorrelation with both actions
Figure 9 shows a typical signal for each food in exper-iment 2 In this experiment we calculate the recognitionaccuracy of food type classification by using Random Forestclassier ofWeka Tables 5 and 6 show the classification resultsThe average recognition accuracy was 896 If we did nothave a SenStick it would be hard to develop the sensorizedchopsticks that can be used in the actual environment
42 Daily Activity Monitoring by a Sensorized Belt Recentlycorpulent people are rapidly increasing and about 700 mil-lion people belong to the obesity group [18] Nowadays manyoffice workers use computers heavily and their sitting time
10 Journal of Sensors
Step 1 crop action Step 2 intake action
Lower
Upper
2 4 6 8 10 12 140Elapsed time (sec)
2 4 6 8 10 12 140Elapsed time (sec)
2 4 6 8 10 12 140Elapsed time (sec)
2 4 6 8 10 12 140Elapsed time (sec)
X
Y
Z
Synthetic
X
Y
Z
Synthetic
X
Y
Z
Synthetic
X
Y
Z
Synthetic
minus250
minus150
minus50
50
150
250
Ang
ular
velo
city
(rad
s)
minus20
minus10
01020
Acce
lera
tion
(ms
2)
minus250
minus150
minus50
50
150
250
Ang
ular
velo
city
(rad
s)
minus20
minus10
01020
Acce
lera
tion
(ms
2)
Open1 Close2 Raise3 Eat4 Drop5
1 2 3 4 5
Figure 8 Time series of sensing data in eating activity
in office increases Keeping incorrect postures during longworking time increases the risk of obesity of office workersTherefore it is necessary tomonitor the daily activity of officeworkers for correcting the bad habits
Motivated by this we have developed a sensorized beltcalled Waiston Belt [8] which aims to monitor a userrsquosdaily activities for maintaining health As products of belt-type wearable devices Welt [19] and Belty [20] are on saleThese provide the function of waist measurement activity
tracking and step counting However API for obtainingsensing data is not provided and it is not suitable for researchpurposes Generally it is hard for ordinary researchers toembed sensors battery BLE and vibrator into a small buckle-size case that can be worn in everyday life Nam et al [21]have proposed a belt-attached device for monitoring postureHowever the device has not reached a sufficient level to beused on a daily basis because of exposed wiring and uncoolappearance However by extending SenStick we succeeded
Journal of Sensors 11
C1
C2 C3
C4
Typical signal of each food
C1 rice amp dish C2 miso soup C3 noodles C4 shrimp
400
200
0
minus200
minus400
Figure 9 Eating activity recognition by sensorized chopsticks
SenStick
Waiston Belt
Figure 10 Waiston Belt based on SenStick
Table 5 Experimental result (food type)
Food type Number ofpieces of data Precision Recall 119865-Measure
C1 rice and dish 3011 0894 0945 0919C2 miso soup 2187 0907 0841 0873C3 noodles 2761 088 0896 0888C4 shrimp 2091 0907 088 0893
0896 0896 0895
to develop a stylish prototype of Waiston Belt shown inFigure 10 quickly Since the battery life of SenStick is 15 hourswith a small battery (60mAh) we can sense continuously inour daily lives As shown in Figure 11 it is possible to estimatevarious usersrsquo contexts (How much did you walk Are youusing the stairs Is posture good at work) by analyzing timeseries of sensing data obtained from the sensorized belt
An important point in this use case is that usersrsquo contextscan be estimated by combining multiple sensing data For
Table 6 Confusion matrix of food type recognition
PredictedC1 C2 C3 C4
ActualC1 rice and dish 094 004 001 000C2 miso soup 012 084 004 000C3 noodles 002 002 090 006C4 shrimp 001 001 011 088
example climbing stairs is estimated from the changes ofacceleration and atmospheric pressure The change of thehumidity data will be helpful to detect that a user enteredan air-conditioned room A SenStick makes it possible toeasily carry out research on behavior recognition with acombination of such multiple sensors
We introduce basic experiments about the posture recog-nition during deskwork by using this beltThe purpose of thisexperiment is to identify the posture type from the attitude
12 Journal of Sensors
Enter the lab
HumidityTemp
Humidity dropping
Pressure dropping
Working
Walking WalkingClimbing stairs Sitting (good posture) Sitting (bad posture)
AccsXAccsYAccsZ
60
40
20
1003
1002
1001
1000
999
25
2
15
1
05
0
minus15
minus1
minus05
Air pressure
Figure 11 Time series of sensing data in daily activity
C1 good posture C2 bad posture 1
C3 bad posture 2 C4 bad posture 3
Figure 12 Time series of sensing data in posture monitoring
angle of the belt calculated from 3 axial acceleration sensorsWe set the sensing rate as 10Hz and collected 6000 samples(10min) for each posture We assumed 4 classes (C1 goodposture C2 bad posture 1 C3 bad posture 2 and C4 badposture 3) as shown in Figure 12
In this experiment we calculated the recognition accu-racy of posture type classification by using Random Forestclassifier of Weka Tables 7 and 8 show the classificationresults The average recognition accuracy was 997 Inthe future based on the result of posture estimation we
Journal of Sensors 13
Sensorized glasses
Figure 13 Sensorized glasses
Table 7 Experimental result (posture type)
Posture type Number ofpieces of data Precision Recall 119865-Measure
C1 good posture 6000 1000 1000 1000C2 bad posture 1 6000 0994 0994 0994C3 bad posture 2 6000 0994 0994 0994C4 bad posture 3 6000 0999 0999 0999
0997 0997 0997
will develop a software application that suggests correctinghisher bad posture to a right posture during desk work
43 Indoor Localization by Sensorized Glasses We have pro-posed an indoor localization system using illuminance-basedtrilateration method [22] In this system we assume that theperson wears glasses having an illuminance sensor as shownin Figure 13 and controllable light sources are deployed in thetarget environment Our system controls lights andmeasuresthe variation of illuminance at the glasses As shown inFigure 14 the distances (1198891ndash1198893) from each light (1198971ndash1198973) to theuser (119906) are estimated from the distance-illumination modelconstructed in advance and userrsquos position is estimated basedon trilateration method using estimated distances
As a result of an experiment where three controllablelight sources were deployed in a room we confirmed thatour system could estimate the position within 05m erroron average By using SenStick we could easily develop thesensorized glasses equipped with illuminance sensor
44 Sensorized Airsoft Gun We have developed the sen-sorized airsoft gun for activity recognition in airsoft gameas shown in Figure 15 Airsoft game is a sport in whichparticipants eliminate opponents by hitting each other usingreplica weapons called airsoft guns The graph of a typicalsignal from the sensorized airsoft gun is shown in Figure 16A user can count the number of shots of a ball by attachingSenStick to his airsoft gun
45 Another Example Studentrsquos Unique Achievements andSurvey Results SenStick is also used in a lecture of IoT
Table 8 Confusion matrix of posture type recognition
PredictedC1 C2 C3 C4
ActualC1 good posture 1000 0000 0000 0000C2 bad posture 1 0000 0994 0006 0001C3 bad posture 2 0000 0006 0994 0000C4 bad posture 3 0000 0001 0000 0999
prototyping Students realize their unique IoT devices easilyby using SenSticks Here we introduce some of the developedIoT devices and survey results about the usefulness ofSenStick answered by students
Figure 17 shows a sensorized pencil with two SenSticksThe sensorized pencil is developed to recognize the charac-ters written by this pencil Even though two SenSticks areembedded to detect precise movement of pencil the size ofthe pencil is reasonable
Figure 18 shows a sensorized name tag anddigital signageSince we changed the battery from 60mAh to 900mAh thetag keeps running more than one week It aims to recordemployeesrsquo or studentsrsquo daily activities to support their healthThe digital signage informs the user who wears a sensorizedname tag about the userrsquos activity data when the user standsin front of it
Figure 19 shows a sensorized spoon used in the digitalfish catching game for children In this game the sensorizedspoon detects the scooping movement of the hand Thuschildren can catch the digital fish displayed by the projectionmapping by using sensorized spoonThis systemwas actuallyused at the universityrsquos open campus event and around100 children enjoyed this game and showed interest in IoTtechnology
The survey results obtained from the ten graduate stu-dents (ninemales and one female) who developed prototypesas mentioned above are shown in Table 9 A questionnaireis a five-grade form (5 strongly agree 4 agree a little 3neither agree nor disagree 2 disagree a little and 1 stronglydisagree) From the results of questions 1 and 2 manyparticipants think that SenStick platform is useful for IoT
14 Journal of Sensors
Position
Distance illumination model
Using trilateration method
a
Light 3
Light 2
Light 1
L = minus421d + 1496
L = minus063d + 501
05 1 15 2 25 3 35 4 450Distance between the illuminance sensor and the lighting device (m)
02468
1012141618
Illum
inan
ce v
alue
(lux
)
Distanceu
d1
d2
d3
l1 p1 x1
l3 p3 x3
l2 p2 x2
Figure 14 Indoor localization by sensorized glasses
Sensorized airsoft gun
Figure 15 Sensorized airsoft gun
research and IoT education Also the results of question 3show that many students are interested in the SenStick andwant to use it again From the results of questions 4 and 5we can see that usability and specification of SenStick arerelatively high There were opinions about the usefulness ofSenStick platform such as ldquoI could quickly try out ideas of IoTprototypingrdquo ldquoIt was easy to collect sensing datardquo and ldquoSinceSenStick is small it was easily attached tomost of the thingsrdquoOn the other hand there were opinions about improvements
such as ldquoI want to use higher sensing frequency (50ndash100Hz)rdquoldquoI want to process a sensing data on SenStickrdquo and ldquoI want tomodify the firmware itself depending on the application ofthe IoT devicerdquo
5 Conclusion
In this paper we have introduced a smart and comprehensivesensing platform called SenStick which is composed of
Journal of Sensors 15
Typical signal of airso gun
Hold a gun Ready
Shoot
1 2 3 4 5 6 70
00
30
60
91
21
51
82
12
42
73
03
33
63
94
24
54
85
15
45
76
06
36
66
97
27
57
88
18
48
79
09
39
69
910
210
510
811
111
411
712
012
312
612
913
2
minus600
minus400
minus200
0
200
400
600
800
acc_xacc_yacc_z
gyro_xgyro_y
acc_synthesis gyro_zgyro_synthesis
Figure 16 Time series of sensing data in shooting an airsoft gun
Sensorized pencil
Figure 17 Sensorized pencil
Sensorized name tagVisualization of sensor data
Figure 18 Sensorized name tag
Sensorized spoon Digital fishing game
Figure 19 Sensorized spoon
Table 9 Questionnaire results about SenStick
Question contents ScoreQ1 Do you think that SenStick will be useful for IoTresearch 49
Q2 Do you think that SenStick will be useful for IoTeducation 47
Q3 Do you want to use SenStick again 49Q4 Do you think that SenStick was easy to use 41Q5 Are you satisfied with specification of SenStick 41
hardware software and 3D case data The main purpose ofour platform is to simplify the sensorization of our personalbelongings
The hardware of SenStick is maybe one of the worldrsquossmallest sensor boards which includes 8 kinds of sensorsand BLE In addition to the hardware we also developedsupport software for managing monitoring recording andviewing the hardware By providing 3D case data all theresearchers can measure every tiny activity in daily life easilyand efficiently For an educational use it is also useful forunderstanding sensors and prototyping the desired systemincluding tiny censored things
Conflicts of Interest
The authors declare that there are no conflicts of interestregarding the publication of this paper
Acknowledgments
This work is supported in part by JST PRESTO and JSPSKAKENHI (Grant no 17J10021) and ldquoResearch and Devel-opment of Innovative Network Technologies to Create theFuturerdquo the Commissioned Research of National Instituteof Information and Communications Technology (NICT)Japan
16 Journal of Sensors
References
[1] O Brdiczka M Langet J Maisonnasse and J L CrowleyldquoDetecting human behavior models from multimodal observa-tion in a smart homerdquo IEEE Transactions on Automation Scienceand Engineering vol 6 no 4 pp 588ndash597 2009
[2] J Chen A Kam J Zhang N Liu and L Shue ldquoBathroomactivity monitoring based on soundrdquo in Pervasive ComputingThird International Conference PERVASIVE 2005 MunichGermanyMay 8-13 2005 Proceedings vol 3468 ofLectureNotesin Computer Science pp 47ndash61 Springer Berlin Germany2002
[3] G Bauer K Stockinger and P Lukowicz ldquoRecognizing theUse-Mode of Kitchen Appliances from Their Current Consump-tionrdquo in Smart Sensing and Context vol 5741 of Lecture Notesin Computer Science pp 163ndash176 Springer Berlin HeidelbergBerlin Heidelberg 2009
[4] K Ouchi and M Doi ldquoIndoor-outdoor activity recognition bya smartphonerdquo in Proceedings of the the 2012 ACM Conferencep 537 Pittsburgh Pennsylvania September 2012
[5] A Anjum and M U Ilyas ldquoActivity recognition using smart-phone sensorsrdquo in Proceedings of the IEEE 10th ConsumerCommunications and Networking Conference (CCNC rsquo13) pp914ndash919 IEEE January 2013
[6] O D Lara and M A Labrador ldquoA survey on human activityrecognition using wearable sensorsrdquo IEEE CommunicationsSurveys amp Tutorials vol 15 no 3 pp 1192ndash1209 2013
[7] ldquoJins memerdquo httpsjins-memecom[8] Y Matsuda T Ishioka T Hasegawa et al ldquoWaistonBelt A belt
for monitoring your real abdominal circumference foreverrdquo inProceedings of the ACM International Joint Conference on Perva-sive and Ubiquitous Computing and the 2015 ACM InternationalSymposium on Wearable Computers UbiComp and ISWC 2015pp 329ndash332 September 2015
[9] M Kurahashi K Murao T Terada and M Tsukamoto ldquoPer-sonal identification system based on rotation of toilet paperrollsrdquo in Proceedings of the IEEE International Conference onPervasive Computing and CommunicationsWorkshops (PerComWorkshops rsquo17) pp 521ndash526 IEEE Kona Hawaii USA March2017
[10] ldquoHapiforkrdquo httpswwwhapicom[11] Y Arakawa ldquoSenStick Sensorize every thingsrdquo in Proceedings of
the ACM International Joint Conference on Pervasive and Ubiq-uitous Computing and the 2015 ACM International SymposiumonWearable Computers UbiComp and ISWC 2015 pp 349ndash352September 2015
[12] Y Nakamura T Kanehira Y Arakawa and K YasumotoldquoSenStick 2 Ultra tiny all-in-one sensor with wireless chargingrdquoinProceedings of the 2016ACM International Joint Conference onPervasive and Ubiquitous Computing UbiComp 2016 pp 337ndash340 September 2016
[13] ldquoSensortag cc2541 (texas instruments)rdquo httpwwwticomtoolcc2541dk-sensor
[14] ldquoIot smart modulerdquo httpwwwalpscomjiotsmart[15] ldquoTsnd121 (atr-promotions)rdquo httpwwwatr-pcomproducts
TSND121html[16] ldquoWax9 (axivity)rdquo httpaxivitycomproductwax9[17] Uk biobank ldquoUk biobankrdquo httpwwwukbiobankacuk[18] J C Seidell and J Halberstadt ldquoThe global burden of obesity
and the challenges of preventionrdquo Annals of Nutrition andMetabolism vol 66 supplement 2 pp 7ndash12 2015
[19] ldquoWeltThe smart belt for fashion and healthrdquo httpswwwwelt-corpcom
[20] ldquoBelty power The belt your really deserverdquo httpswwwbeltyco
[21] H Nam J-H Kim and J-I Kim ldquoSmart Belt A wearabledevice for managing abdominal obesityrdquo in Proceedings of theInternational Conference on Big Data and Smart ComputingBigComp 2016 pp 430ndash434 January 2016
[22] K Moriya M Fujimoto Y Arakawa and K Yasumoto ldquoIndoorlocalization based on distance-illuminance model and activecontrol of lighting devicesrdquo in Proceedings of the 2016 Interna-tional Conference on Indoor Positioning and Indoor NavigationIPIN 2016 pp 1ndash6 October 2016
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal of
Volume 201
Submit your manuscripts athttpswwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
Journal of Sensors 9
Chopsticks Tooth brush Glasses Umbrella Name card Cookware
A user can design original case data in a few minutes
Basic 3D case of SenStick
Figure 6 3D case data of SenStick
Sensorized
Chopsticks
Figure 7 Eating activity recognition by sensorized chopsticks
kitchen utensils animals stuffed toy a remote control andany kind of things that can be sensing targets In this sectionwe introduce some case studies results utilizing SenStick
41 Eating Behavior Recognition by Sensorized Chopsticks Inmany cases East Asian people eat meals with chopsticksBy analyzing the data obtained from chopsticks there is apossibility that we can estimate the eating habits includinganswers for questions like ldquohow did you eatrdquo and ldquowhatdid you eatrdquo and so forth These data are very useful formaking healthcare applications and leading a healthy lifeThis is one reason why we develop the sensorized chopsticksusing SenStick shown in Figure 7
Research goals of the eating behavior recognition by sen-sorized chopsticks are as follows G1 (behavior) recognizingthe timing and speed and detecting ldquocroprdquo and ldquointakerdquoaction G2 (food type) classifying food content like ricenoodles and so forth from the eating behavior and G3(person) identifying an individual from using chopsticks Inthis paper we show the research results on G1 and G2
We conducted basic experiments to recognize the eatingbehavior (G1 and G2) We used acceleration and gyroscopedata measured by SenStick where the sensing rate is 10HzThe purpose of experiment 1 is to understand the basic
characteristics of sensing data from sensorized chopstickwhen a participant ate chips The purpose of experiment 2 isto identify the food type from the difference in movement ofchopsticks We assumed 4 classes (C1 rice and dish C2 misosoup C3 noodles and C4 shrimp) as food types
Figure 8 shows time series of sensing data for a set ofeating actions in experiment 1 The eating actions are cropaction and intake action In crop action 6-axis motionsensors detect movement (open and close) of the chopsticksIn intake action sensors detect movement of the arm fromthe dish to the mouth Each graph in Figure 8 shows certaincorrelation with both actions
Figure 9 shows a typical signal for each food in exper-iment 2 In this experiment we calculate the recognitionaccuracy of food type classification by using Random Forestclassier ofWeka Tables 5 and 6 show the classification resultsThe average recognition accuracy was 896 If we did nothave a SenStick it would be hard to develop the sensorizedchopsticks that can be used in the actual environment
42 Daily Activity Monitoring by a Sensorized Belt Recentlycorpulent people are rapidly increasing and about 700 mil-lion people belong to the obesity group [18] Nowadays manyoffice workers use computers heavily and their sitting time
10 Journal of Sensors
Step 1 crop action Step 2 intake action
Lower
Upper
2 4 6 8 10 12 140Elapsed time (sec)
2 4 6 8 10 12 140Elapsed time (sec)
2 4 6 8 10 12 140Elapsed time (sec)
2 4 6 8 10 12 140Elapsed time (sec)
X
Y
Z
Synthetic
X
Y
Z
Synthetic
X
Y
Z
Synthetic
X
Y
Z
Synthetic
minus250
minus150
minus50
50
150
250
Ang
ular
velo
city
(rad
s)
minus20
minus10
01020
Acce
lera
tion
(ms
2)
minus250
minus150
minus50
50
150
250
Ang
ular
velo
city
(rad
s)
minus20
minus10
01020
Acce
lera
tion
(ms
2)
Open1 Close2 Raise3 Eat4 Drop5
1 2 3 4 5
Figure 8 Time series of sensing data in eating activity
in office increases Keeping incorrect postures during longworking time increases the risk of obesity of office workersTherefore it is necessary tomonitor the daily activity of officeworkers for correcting the bad habits
Motivated by this we have developed a sensorized beltcalled Waiston Belt [8] which aims to monitor a userrsquosdaily activities for maintaining health As products of belt-type wearable devices Welt [19] and Belty [20] are on saleThese provide the function of waist measurement activity
tracking and step counting However API for obtainingsensing data is not provided and it is not suitable for researchpurposes Generally it is hard for ordinary researchers toembed sensors battery BLE and vibrator into a small buckle-size case that can be worn in everyday life Nam et al [21]have proposed a belt-attached device for monitoring postureHowever the device has not reached a sufficient level to beused on a daily basis because of exposed wiring and uncoolappearance However by extending SenStick we succeeded
Journal of Sensors 11
C1
C2 C3
C4
Typical signal of each food
C1 rice amp dish C2 miso soup C3 noodles C4 shrimp
400
200
0
minus200
minus400
Figure 9 Eating activity recognition by sensorized chopsticks
SenStick
Waiston Belt
Figure 10 Waiston Belt based on SenStick
Table 5 Experimental result (food type)
Food type Number ofpieces of data Precision Recall 119865-Measure
C1 rice and dish 3011 0894 0945 0919C2 miso soup 2187 0907 0841 0873C3 noodles 2761 088 0896 0888C4 shrimp 2091 0907 088 0893
0896 0896 0895
to develop a stylish prototype of Waiston Belt shown inFigure 10 quickly Since the battery life of SenStick is 15 hourswith a small battery (60mAh) we can sense continuously inour daily lives As shown in Figure 11 it is possible to estimatevarious usersrsquo contexts (How much did you walk Are youusing the stairs Is posture good at work) by analyzing timeseries of sensing data obtained from the sensorized belt
An important point in this use case is that usersrsquo contextscan be estimated by combining multiple sensing data For
Table 6 Confusion matrix of food type recognition
PredictedC1 C2 C3 C4
ActualC1 rice and dish 094 004 001 000C2 miso soup 012 084 004 000C3 noodles 002 002 090 006C4 shrimp 001 001 011 088
example climbing stairs is estimated from the changes ofacceleration and atmospheric pressure The change of thehumidity data will be helpful to detect that a user enteredan air-conditioned room A SenStick makes it possible toeasily carry out research on behavior recognition with acombination of such multiple sensors
We introduce basic experiments about the posture recog-nition during deskwork by using this beltThe purpose of thisexperiment is to identify the posture type from the attitude
12 Journal of Sensors
Enter the lab
HumidityTemp
Humidity dropping
Pressure dropping
Working
Walking WalkingClimbing stairs Sitting (good posture) Sitting (bad posture)
AccsXAccsYAccsZ
60
40
20
1003
1002
1001
1000
999
25
2
15
1
05
0
minus15
minus1
minus05
Air pressure
Figure 11 Time series of sensing data in daily activity
C1 good posture C2 bad posture 1
C3 bad posture 2 C4 bad posture 3
Figure 12 Time series of sensing data in posture monitoring
angle of the belt calculated from 3 axial acceleration sensorsWe set the sensing rate as 10Hz and collected 6000 samples(10min) for each posture We assumed 4 classes (C1 goodposture C2 bad posture 1 C3 bad posture 2 and C4 badposture 3) as shown in Figure 12
In this experiment we calculated the recognition accu-racy of posture type classification by using Random Forestclassifier of Weka Tables 7 and 8 show the classificationresults The average recognition accuracy was 997 Inthe future based on the result of posture estimation we
Journal of Sensors 13
Sensorized glasses
Figure 13 Sensorized glasses
Table 7 Experimental result (posture type)
Posture type Number ofpieces of data Precision Recall 119865-Measure
C1 good posture 6000 1000 1000 1000C2 bad posture 1 6000 0994 0994 0994C3 bad posture 2 6000 0994 0994 0994C4 bad posture 3 6000 0999 0999 0999
0997 0997 0997
will develop a software application that suggests correctinghisher bad posture to a right posture during desk work
43 Indoor Localization by Sensorized Glasses We have pro-posed an indoor localization system using illuminance-basedtrilateration method [22] In this system we assume that theperson wears glasses having an illuminance sensor as shownin Figure 13 and controllable light sources are deployed in thetarget environment Our system controls lights andmeasuresthe variation of illuminance at the glasses As shown inFigure 14 the distances (1198891ndash1198893) from each light (1198971ndash1198973) to theuser (119906) are estimated from the distance-illumination modelconstructed in advance and userrsquos position is estimated basedon trilateration method using estimated distances
As a result of an experiment where three controllablelight sources were deployed in a room we confirmed thatour system could estimate the position within 05m erroron average By using SenStick we could easily develop thesensorized glasses equipped with illuminance sensor
44 Sensorized Airsoft Gun We have developed the sen-sorized airsoft gun for activity recognition in airsoft gameas shown in Figure 15 Airsoft game is a sport in whichparticipants eliminate opponents by hitting each other usingreplica weapons called airsoft guns The graph of a typicalsignal from the sensorized airsoft gun is shown in Figure 16A user can count the number of shots of a ball by attachingSenStick to his airsoft gun
45 Another Example Studentrsquos Unique Achievements andSurvey Results SenStick is also used in a lecture of IoT
Table 8 Confusion matrix of posture type recognition
PredictedC1 C2 C3 C4
ActualC1 good posture 1000 0000 0000 0000C2 bad posture 1 0000 0994 0006 0001C3 bad posture 2 0000 0006 0994 0000C4 bad posture 3 0000 0001 0000 0999
prototyping Students realize their unique IoT devices easilyby using SenSticks Here we introduce some of the developedIoT devices and survey results about the usefulness ofSenStick answered by students
Figure 17 shows a sensorized pencil with two SenSticksThe sensorized pencil is developed to recognize the charac-ters written by this pencil Even though two SenSticks areembedded to detect precise movement of pencil the size ofthe pencil is reasonable
Figure 18 shows a sensorized name tag anddigital signageSince we changed the battery from 60mAh to 900mAh thetag keeps running more than one week It aims to recordemployeesrsquo or studentsrsquo daily activities to support their healthThe digital signage informs the user who wears a sensorizedname tag about the userrsquos activity data when the user standsin front of it
Figure 19 shows a sensorized spoon used in the digitalfish catching game for children In this game the sensorizedspoon detects the scooping movement of the hand Thuschildren can catch the digital fish displayed by the projectionmapping by using sensorized spoonThis systemwas actuallyused at the universityrsquos open campus event and around100 children enjoyed this game and showed interest in IoTtechnology
The survey results obtained from the ten graduate stu-dents (ninemales and one female) who developed prototypesas mentioned above are shown in Table 9 A questionnaireis a five-grade form (5 strongly agree 4 agree a little 3neither agree nor disagree 2 disagree a little and 1 stronglydisagree) From the results of questions 1 and 2 manyparticipants think that SenStick platform is useful for IoT
14 Journal of Sensors
Position
Distance illumination model
Using trilateration method
a
Light 3
Light 2
Light 1
L = minus421d + 1496
L = minus063d + 501
05 1 15 2 25 3 35 4 450Distance between the illuminance sensor and the lighting device (m)
02468
1012141618
Illum
inan
ce v
alue
(lux
)
Distanceu
d1
d2
d3
l1 p1 x1
l3 p3 x3
l2 p2 x2
Figure 14 Indoor localization by sensorized glasses
Sensorized airsoft gun
Figure 15 Sensorized airsoft gun
research and IoT education Also the results of question 3show that many students are interested in the SenStick andwant to use it again From the results of questions 4 and 5we can see that usability and specification of SenStick arerelatively high There were opinions about the usefulness ofSenStick platform such as ldquoI could quickly try out ideas of IoTprototypingrdquo ldquoIt was easy to collect sensing datardquo and ldquoSinceSenStick is small it was easily attached tomost of the thingsrdquoOn the other hand there were opinions about improvements
such as ldquoI want to use higher sensing frequency (50ndash100Hz)rdquoldquoI want to process a sensing data on SenStickrdquo and ldquoI want tomodify the firmware itself depending on the application ofthe IoT devicerdquo
5 Conclusion
In this paper we have introduced a smart and comprehensivesensing platform called SenStick which is composed of
Journal of Sensors 15
Typical signal of airso gun
Hold a gun Ready
Shoot
1 2 3 4 5 6 70
00
30
60
91
21
51
82
12
42
73
03
33
63
94
24
54
85
15
45
76
06
36
66
97
27
57
88
18
48
79
09
39
69
910
210
510
811
111
411
712
012
312
612
913
2
minus600
minus400
minus200
0
200
400
600
800
acc_xacc_yacc_z
gyro_xgyro_y
acc_synthesis gyro_zgyro_synthesis
Figure 16 Time series of sensing data in shooting an airsoft gun
Sensorized pencil
Figure 17 Sensorized pencil
Sensorized name tagVisualization of sensor data
Figure 18 Sensorized name tag
Sensorized spoon Digital fishing game
Figure 19 Sensorized spoon
Table 9 Questionnaire results about SenStick
Question contents ScoreQ1 Do you think that SenStick will be useful for IoTresearch 49
Q2 Do you think that SenStick will be useful for IoTeducation 47
Q3 Do you want to use SenStick again 49Q4 Do you think that SenStick was easy to use 41Q5 Are you satisfied with specification of SenStick 41
hardware software and 3D case data The main purpose ofour platform is to simplify the sensorization of our personalbelongings
The hardware of SenStick is maybe one of the worldrsquossmallest sensor boards which includes 8 kinds of sensorsand BLE In addition to the hardware we also developedsupport software for managing monitoring recording andviewing the hardware By providing 3D case data all theresearchers can measure every tiny activity in daily life easilyand efficiently For an educational use it is also useful forunderstanding sensors and prototyping the desired systemincluding tiny censored things
Conflicts of Interest
The authors declare that there are no conflicts of interestregarding the publication of this paper
Acknowledgments
This work is supported in part by JST PRESTO and JSPSKAKENHI (Grant no 17J10021) and ldquoResearch and Devel-opment of Innovative Network Technologies to Create theFuturerdquo the Commissioned Research of National Instituteof Information and Communications Technology (NICT)Japan
16 Journal of Sensors
References
[1] O Brdiczka M Langet J Maisonnasse and J L CrowleyldquoDetecting human behavior models from multimodal observa-tion in a smart homerdquo IEEE Transactions on Automation Scienceand Engineering vol 6 no 4 pp 588ndash597 2009
[2] J Chen A Kam J Zhang N Liu and L Shue ldquoBathroomactivity monitoring based on soundrdquo in Pervasive ComputingThird International Conference PERVASIVE 2005 MunichGermanyMay 8-13 2005 Proceedings vol 3468 ofLectureNotesin Computer Science pp 47ndash61 Springer Berlin Germany2002
[3] G Bauer K Stockinger and P Lukowicz ldquoRecognizing theUse-Mode of Kitchen Appliances from Their Current Consump-tionrdquo in Smart Sensing and Context vol 5741 of Lecture Notesin Computer Science pp 163ndash176 Springer Berlin HeidelbergBerlin Heidelberg 2009
[4] K Ouchi and M Doi ldquoIndoor-outdoor activity recognition bya smartphonerdquo in Proceedings of the the 2012 ACM Conferencep 537 Pittsburgh Pennsylvania September 2012
[5] A Anjum and M U Ilyas ldquoActivity recognition using smart-phone sensorsrdquo in Proceedings of the IEEE 10th ConsumerCommunications and Networking Conference (CCNC rsquo13) pp914ndash919 IEEE January 2013
[6] O D Lara and M A Labrador ldquoA survey on human activityrecognition using wearable sensorsrdquo IEEE CommunicationsSurveys amp Tutorials vol 15 no 3 pp 1192ndash1209 2013
[7] ldquoJins memerdquo httpsjins-memecom[8] Y Matsuda T Ishioka T Hasegawa et al ldquoWaistonBelt A belt
for monitoring your real abdominal circumference foreverrdquo inProceedings of the ACM International Joint Conference on Perva-sive and Ubiquitous Computing and the 2015 ACM InternationalSymposium on Wearable Computers UbiComp and ISWC 2015pp 329ndash332 September 2015
[9] M Kurahashi K Murao T Terada and M Tsukamoto ldquoPer-sonal identification system based on rotation of toilet paperrollsrdquo in Proceedings of the IEEE International Conference onPervasive Computing and CommunicationsWorkshops (PerComWorkshops rsquo17) pp 521ndash526 IEEE Kona Hawaii USA March2017
[10] ldquoHapiforkrdquo httpswwwhapicom[11] Y Arakawa ldquoSenStick Sensorize every thingsrdquo in Proceedings of
the ACM International Joint Conference on Pervasive and Ubiq-uitous Computing and the 2015 ACM International SymposiumonWearable Computers UbiComp and ISWC 2015 pp 349ndash352September 2015
[12] Y Nakamura T Kanehira Y Arakawa and K YasumotoldquoSenStick 2 Ultra tiny all-in-one sensor with wireless chargingrdquoinProceedings of the 2016ACM International Joint Conference onPervasive and Ubiquitous Computing UbiComp 2016 pp 337ndash340 September 2016
[13] ldquoSensortag cc2541 (texas instruments)rdquo httpwwwticomtoolcc2541dk-sensor
[14] ldquoIot smart modulerdquo httpwwwalpscomjiotsmart[15] ldquoTsnd121 (atr-promotions)rdquo httpwwwatr-pcomproducts
TSND121html[16] ldquoWax9 (axivity)rdquo httpaxivitycomproductwax9[17] Uk biobank ldquoUk biobankrdquo httpwwwukbiobankacuk[18] J C Seidell and J Halberstadt ldquoThe global burden of obesity
and the challenges of preventionrdquo Annals of Nutrition andMetabolism vol 66 supplement 2 pp 7ndash12 2015
[19] ldquoWeltThe smart belt for fashion and healthrdquo httpswwwwelt-corpcom
[20] ldquoBelty power The belt your really deserverdquo httpswwwbeltyco
[21] H Nam J-H Kim and J-I Kim ldquoSmart Belt A wearabledevice for managing abdominal obesityrdquo in Proceedings of theInternational Conference on Big Data and Smart ComputingBigComp 2016 pp 430ndash434 January 2016
[22] K Moriya M Fujimoto Y Arakawa and K Yasumoto ldquoIndoorlocalization based on distance-illuminance model and activecontrol of lighting devicesrdquo in Proceedings of the 2016 Interna-tional Conference on Indoor Positioning and Indoor NavigationIPIN 2016 pp 1ndash6 October 2016
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal of
Volume 201
Submit your manuscripts athttpswwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
10 Journal of Sensors
Step 1 crop action Step 2 intake action
Lower
Upper
2 4 6 8 10 12 140Elapsed time (sec)
2 4 6 8 10 12 140Elapsed time (sec)
2 4 6 8 10 12 140Elapsed time (sec)
2 4 6 8 10 12 140Elapsed time (sec)
X
Y
Z
Synthetic
X
Y
Z
Synthetic
X
Y
Z
Synthetic
X
Y
Z
Synthetic
minus250
minus150
minus50
50
150
250
Ang
ular
velo
city
(rad
s)
minus20
minus10
01020
Acce
lera
tion
(ms
2)
minus250
minus150
minus50
50
150
250
Ang
ular
velo
city
(rad
s)
minus20
minus10
01020
Acce
lera
tion
(ms
2)
Open1 Close2 Raise3 Eat4 Drop5
1 2 3 4 5
Figure 8 Time series of sensing data in eating activity
in office increases Keeping incorrect postures during longworking time increases the risk of obesity of office workersTherefore it is necessary tomonitor the daily activity of officeworkers for correcting the bad habits
Motivated by this we have developed a sensorized beltcalled Waiston Belt [8] which aims to monitor a userrsquosdaily activities for maintaining health As products of belt-type wearable devices Welt [19] and Belty [20] are on saleThese provide the function of waist measurement activity
tracking and step counting However API for obtainingsensing data is not provided and it is not suitable for researchpurposes Generally it is hard for ordinary researchers toembed sensors battery BLE and vibrator into a small buckle-size case that can be worn in everyday life Nam et al [21]have proposed a belt-attached device for monitoring postureHowever the device has not reached a sufficient level to beused on a daily basis because of exposed wiring and uncoolappearance However by extending SenStick we succeeded
Journal of Sensors 11
C1
C2 C3
C4
Typical signal of each food
C1 rice amp dish C2 miso soup C3 noodles C4 shrimp
400
200
0
minus200
minus400
Figure 9 Eating activity recognition by sensorized chopsticks
SenStick
Waiston Belt
Figure 10 Waiston Belt based on SenStick
Table 5 Experimental result (food type)
Food type Number ofpieces of data Precision Recall 119865-Measure
C1 rice and dish 3011 0894 0945 0919C2 miso soup 2187 0907 0841 0873C3 noodles 2761 088 0896 0888C4 shrimp 2091 0907 088 0893
0896 0896 0895
to develop a stylish prototype of Waiston Belt shown inFigure 10 quickly Since the battery life of SenStick is 15 hourswith a small battery (60mAh) we can sense continuously inour daily lives As shown in Figure 11 it is possible to estimatevarious usersrsquo contexts (How much did you walk Are youusing the stairs Is posture good at work) by analyzing timeseries of sensing data obtained from the sensorized belt
An important point in this use case is that usersrsquo contextscan be estimated by combining multiple sensing data For
Table 6 Confusion matrix of food type recognition
PredictedC1 C2 C3 C4
ActualC1 rice and dish 094 004 001 000C2 miso soup 012 084 004 000C3 noodles 002 002 090 006C4 shrimp 001 001 011 088
example climbing stairs is estimated from the changes ofacceleration and atmospheric pressure The change of thehumidity data will be helpful to detect that a user enteredan air-conditioned room A SenStick makes it possible toeasily carry out research on behavior recognition with acombination of such multiple sensors
We introduce basic experiments about the posture recog-nition during deskwork by using this beltThe purpose of thisexperiment is to identify the posture type from the attitude
12 Journal of Sensors
Enter the lab
HumidityTemp
Humidity dropping
Pressure dropping
Working
Walking WalkingClimbing stairs Sitting (good posture) Sitting (bad posture)
AccsXAccsYAccsZ
60
40
20
1003
1002
1001
1000
999
25
2
15
1
05
0
minus15
minus1
minus05
Air pressure
Figure 11 Time series of sensing data in daily activity
C1 good posture C2 bad posture 1
C3 bad posture 2 C4 bad posture 3
Figure 12 Time series of sensing data in posture monitoring
angle of the belt calculated from 3 axial acceleration sensorsWe set the sensing rate as 10Hz and collected 6000 samples(10min) for each posture We assumed 4 classes (C1 goodposture C2 bad posture 1 C3 bad posture 2 and C4 badposture 3) as shown in Figure 12
In this experiment we calculated the recognition accu-racy of posture type classification by using Random Forestclassifier of Weka Tables 7 and 8 show the classificationresults The average recognition accuracy was 997 Inthe future based on the result of posture estimation we
Journal of Sensors 13
Sensorized glasses
Figure 13 Sensorized glasses
Table 7 Experimental result (posture type)
Posture type Number ofpieces of data Precision Recall 119865-Measure
C1 good posture 6000 1000 1000 1000C2 bad posture 1 6000 0994 0994 0994C3 bad posture 2 6000 0994 0994 0994C4 bad posture 3 6000 0999 0999 0999
0997 0997 0997
will develop a software application that suggests correctinghisher bad posture to a right posture during desk work
43 Indoor Localization by Sensorized Glasses We have pro-posed an indoor localization system using illuminance-basedtrilateration method [22] In this system we assume that theperson wears glasses having an illuminance sensor as shownin Figure 13 and controllable light sources are deployed in thetarget environment Our system controls lights andmeasuresthe variation of illuminance at the glasses As shown inFigure 14 the distances (1198891ndash1198893) from each light (1198971ndash1198973) to theuser (119906) are estimated from the distance-illumination modelconstructed in advance and userrsquos position is estimated basedon trilateration method using estimated distances
As a result of an experiment where three controllablelight sources were deployed in a room we confirmed thatour system could estimate the position within 05m erroron average By using SenStick we could easily develop thesensorized glasses equipped with illuminance sensor
44 Sensorized Airsoft Gun We have developed the sen-sorized airsoft gun for activity recognition in airsoft gameas shown in Figure 15 Airsoft game is a sport in whichparticipants eliminate opponents by hitting each other usingreplica weapons called airsoft guns The graph of a typicalsignal from the sensorized airsoft gun is shown in Figure 16A user can count the number of shots of a ball by attachingSenStick to his airsoft gun
45 Another Example Studentrsquos Unique Achievements andSurvey Results SenStick is also used in a lecture of IoT
Table 8 Confusion matrix of posture type recognition
PredictedC1 C2 C3 C4
ActualC1 good posture 1000 0000 0000 0000C2 bad posture 1 0000 0994 0006 0001C3 bad posture 2 0000 0006 0994 0000C4 bad posture 3 0000 0001 0000 0999
prototyping Students realize their unique IoT devices easilyby using SenSticks Here we introduce some of the developedIoT devices and survey results about the usefulness ofSenStick answered by students
Figure 17 shows a sensorized pencil with two SenSticksThe sensorized pencil is developed to recognize the charac-ters written by this pencil Even though two SenSticks areembedded to detect precise movement of pencil the size ofthe pencil is reasonable
Figure 18 shows a sensorized name tag anddigital signageSince we changed the battery from 60mAh to 900mAh thetag keeps running more than one week It aims to recordemployeesrsquo or studentsrsquo daily activities to support their healthThe digital signage informs the user who wears a sensorizedname tag about the userrsquos activity data when the user standsin front of it
Figure 19 shows a sensorized spoon used in the digitalfish catching game for children In this game the sensorizedspoon detects the scooping movement of the hand Thuschildren can catch the digital fish displayed by the projectionmapping by using sensorized spoonThis systemwas actuallyused at the universityrsquos open campus event and around100 children enjoyed this game and showed interest in IoTtechnology
The survey results obtained from the ten graduate stu-dents (ninemales and one female) who developed prototypesas mentioned above are shown in Table 9 A questionnaireis a five-grade form (5 strongly agree 4 agree a little 3neither agree nor disagree 2 disagree a little and 1 stronglydisagree) From the results of questions 1 and 2 manyparticipants think that SenStick platform is useful for IoT
14 Journal of Sensors
Position
Distance illumination model
Using trilateration method
a
Light 3
Light 2
Light 1
L = minus421d + 1496
L = minus063d + 501
05 1 15 2 25 3 35 4 450Distance between the illuminance sensor and the lighting device (m)
02468
1012141618
Illum
inan
ce v
alue
(lux
)
Distanceu
d1
d2
d3
l1 p1 x1
l3 p3 x3
l2 p2 x2
Figure 14 Indoor localization by sensorized glasses
Sensorized airsoft gun
Figure 15 Sensorized airsoft gun
research and IoT education Also the results of question 3show that many students are interested in the SenStick andwant to use it again From the results of questions 4 and 5we can see that usability and specification of SenStick arerelatively high There were opinions about the usefulness ofSenStick platform such as ldquoI could quickly try out ideas of IoTprototypingrdquo ldquoIt was easy to collect sensing datardquo and ldquoSinceSenStick is small it was easily attached tomost of the thingsrdquoOn the other hand there were opinions about improvements
such as ldquoI want to use higher sensing frequency (50ndash100Hz)rdquoldquoI want to process a sensing data on SenStickrdquo and ldquoI want tomodify the firmware itself depending on the application ofthe IoT devicerdquo
5 Conclusion
In this paper we have introduced a smart and comprehensivesensing platform called SenStick which is composed of
Journal of Sensors 15
Typical signal of airso gun
Hold a gun Ready
Shoot
1 2 3 4 5 6 70
00
30
60
91
21
51
82
12
42
73
03
33
63
94
24
54
85
15
45
76
06
36
66
97
27
57
88
18
48
79
09
39
69
910
210
510
811
111
411
712
012
312
612
913
2
minus600
minus400
minus200
0
200
400
600
800
acc_xacc_yacc_z
gyro_xgyro_y
acc_synthesis gyro_zgyro_synthesis
Figure 16 Time series of sensing data in shooting an airsoft gun
Sensorized pencil
Figure 17 Sensorized pencil
Sensorized name tagVisualization of sensor data
Figure 18 Sensorized name tag
Sensorized spoon Digital fishing game
Figure 19 Sensorized spoon
Table 9 Questionnaire results about SenStick
Question contents ScoreQ1 Do you think that SenStick will be useful for IoTresearch 49
Q2 Do you think that SenStick will be useful for IoTeducation 47
Q3 Do you want to use SenStick again 49Q4 Do you think that SenStick was easy to use 41Q5 Are you satisfied with specification of SenStick 41
hardware software and 3D case data The main purpose ofour platform is to simplify the sensorization of our personalbelongings
The hardware of SenStick is maybe one of the worldrsquossmallest sensor boards which includes 8 kinds of sensorsand BLE In addition to the hardware we also developedsupport software for managing monitoring recording andviewing the hardware By providing 3D case data all theresearchers can measure every tiny activity in daily life easilyand efficiently For an educational use it is also useful forunderstanding sensors and prototyping the desired systemincluding tiny censored things
Conflicts of Interest
The authors declare that there are no conflicts of interestregarding the publication of this paper
Acknowledgments
This work is supported in part by JST PRESTO and JSPSKAKENHI (Grant no 17J10021) and ldquoResearch and Devel-opment of Innovative Network Technologies to Create theFuturerdquo the Commissioned Research of National Instituteof Information and Communications Technology (NICT)Japan
16 Journal of Sensors
References
[1] O Brdiczka M Langet J Maisonnasse and J L CrowleyldquoDetecting human behavior models from multimodal observa-tion in a smart homerdquo IEEE Transactions on Automation Scienceand Engineering vol 6 no 4 pp 588ndash597 2009
[2] J Chen A Kam J Zhang N Liu and L Shue ldquoBathroomactivity monitoring based on soundrdquo in Pervasive ComputingThird International Conference PERVASIVE 2005 MunichGermanyMay 8-13 2005 Proceedings vol 3468 ofLectureNotesin Computer Science pp 47ndash61 Springer Berlin Germany2002
[3] G Bauer K Stockinger and P Lukowicz ldquoRecognizing theUse-Mode of Kitchen Appliances from Their Current Consump-tionrdquo in Smart Sensing and Context vol 5741 of Lecture Notesin Computer Science pp 163ndash176 Springer Berlin HeidelbergBerlin Heidelberg 2009
[4] K Ouchi and M Doi ldquoIndoor-outdoor activity recognition bya smartphonerdquo in Proceedings of the the 2012 ACM Conferencep 537 Pittsburgh Pennsylvania September 2012
[5] A Anjum and M U Ilyas ldquoActivity recognition using smart-phone sensorsrdquo in Proceedings of the IEEE 10th ConsumerCommunications and Networking Conference (CCNC rsquo13) pp914ndash919 IEEE January 2013
[6] O D Lara and M A Labrador ldquoA survey on human activityrecognition using wearable sensorsrdquo IEEE CommunicationsSurveys amp Tutorials vol 15 no 3 pp 1192ndash1209 2013
[7] ldquoJins memerdquo httpsjins-memecom[8] Y Matsuda T Ishioka T Hasegawa et al ldquoWaistonBelt A belt
for monitoring your real abdominal circumference foreverrdquo inProceedings of the ACM International Joint Conference on Perva-sive and Ubiquitous Computing and the 2015 ACM InternationalSymposium on Wearable Computers UbiComp and ISWC 2015pp 329ndash332 September 2015
[9] M Kurahashi K Murao T Terada and M Tsukamoto ldquoPer-sonal identification system based on rotation of toilet paperrollsrdquo in Proceedings of the IEEE International Conference onPervasive Computing and CommunicationsWorkshops (PerComWorkshops rsquo17) pp 521ndash526 IEEE Kona Hawaii USA March2017
[10] ldquoHapiforkrdquo httpswwwhapicom[11] Y Arakawa ldquoSenStick Sensorize every thingsrdquo in Proceedings of
the ACM International Joint Conference on Pervasive and Ubiq-uitous Computing and the 2015 ACM International SymposiumonWearable Computers UbiComp and ISWC 2015 pp 349ndash352September 2015
[12] Y Nakamura T Kanehira Y Arakawa and K YasumotoldquoSenStick 2 Ultra tiny all-in-one sensor with wireless chargingrdquoinProceedings of the 2016ACM International Joint Conference onPervasive and Ubiquitous Computing UbiComp 2016 pp 337ndash340 September 2016
[13] ldquoSensortag cc2541 (texas instruments)rdquo httpwwwticomtoolcc2541dk-sensor
[14] ldquoIot smart modulerdquo httpwwwalpscomjiotsmart[15] ldquoTsnd121 (atr-promotions)rdquo httpwwwatr-pcomproducts
TSND121html[16] ldquoWax9 (axivity)rdquo httpaxivitycomproductwax9[17] Uk biobank ldquoUk biobankrdquo httpwwwukbiobankacuk[18] J C Seidell and J Halberstadt ldquoThe global burden of obesity
and the challenges of preventionrdquo Annals of Nutrition andMetabolism vol 66 supplement 2 pp 7ndash12 2015
[19] ldquoWeltThe smart belt for fashion and healthrdquo httpswwwwelt-corpcom
[20] ldquoBelty power The belt your really deserverdquo httpswwwbeltyco
[21] H Nam J-H Kim and J-I Kim ldquoSmart Belt A wearabledevice for managing abdominal obesityrdquo in Proceedings of theInternational Conference on Big Data and Smart ComputingBigComp 2016 pp 430ndash434 January 2016
[22] K Moriya M Fujimoto Y Arakawa and K Yasumoto ldquoIndoorlocalization based on distance-illuminance model and activecontrol of lighting devicesrdquo in Proceedings of the 2016 Interna-tional Conference on Indoor Positioning and Indoor NavigationIPIN 2016 pp 1ndash6 October 2016
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal of
Volume 201
Submit your manuscripts athttpswwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
Journal of Sensors 11
C1
C2 C3
C4
Typical signal of each food
C1 rice amp dish C2 miso soup C3 noodles C4 shrimp
400
200
0
minus200
minus400
Figure 9 Eating activity recognition by sensorized chopsticks
SenStick
Waiston Belt
Figure 10 Waiston Belt based on SenStick
Table 5 Experimental result (food type)
Food type Number ofpieces of data Precision Recall 119865-Measure
C1 rice and dish 3011 0894 0945 0919C2 miso soup 2187 0907 0841 0873C3 noodles 2761 088 0896 0888C4 shrimp 2091 0907 088 0893
0896 0896 0895
to develop a stylish prototype of Waiston Belt shown inFigure 10 quickly Since the battery life of SenStick is 15 hourswith a small battery (60mAh) we can sense continuously inour daily lives As shown in Figure 11 it is possible to estimatevarious usersrsquo contexts (How much did you walk Are youusing the stairs Is posture good at work) by analyzing timeseries of sensing data obtained from the sensorized belt
An important point in this use case is that usersrsquo contextscan be estimated by combining multiple sensing data For
Table 6 Confusion matrix of food type recognition
PredictedC1 C2 C3 C4
ActualC1 rice and dish 094 004 001 000C2 miso soup 012 084 004 000C3 noodles 002 002 090 006C4 shrimp 001 001 011 088
example climbing stairs is estimated from the changes ofacceleration and atmospheric pressure The change of thehumidity data will be helpful to detect that a user enteredan air-conditioned room A SenStick makes it possible toeasily carry out research on behavior recognition with acombination of such multiple sensors
We introduce basic experiments about the posture recog-nition during deskwork by using this beltThe purpose of thisexperiment is to identify the posture type from the attitude
12 Journal of Sensors
Enter the lab
HumidityTemp
Humidity dropping
Pressure dropping
Working
Walking WalkingClimbing stairs Sitting (good posture) Sitting (bad posture)
AccsXAccsYAccsZ
60
40
20
1003
1002
1001
1000
999
25
2
15
1
05
0
minus15
minus1
minus05
Air pressure
Figure 11 Time series of sensing data in daily activity
C1 good posture C2 bad posture 1
C3 bad posture 2 C4 bad posture 3
Figure 12 Time series of sensing data in posture monitoring
angle of the belt calculated from 3 axial acceleration sensorsWe set the sensing rate as 10Hz and collected 6000 samples(10min) for each posture We assumed 4 classes (C1 goodposture C2 bad posture 1 C3 bad posture 2 and C4 badposture 3) as shown in Figure 12
In this experiment we calculated the recognition accu-racy of posture type classification by using Random Forestclassifier of Weka Tables 7 and 8 show the classificationresults The average recognition accuracy was 997 Inthe future based on the result of posture estimation we
Journal of Sensors 13
Sensorized glasses
Figure 13 Sensorized glasses
Table 7 Experimental result (posture type)
Posture type Number ofpieces of data Precision Recall 119865-Measure
C1 good posture 6000 1000 1000 1000C2 bad posture 1 6000 0994 0994 0994C3 bad posture 2 6000 0994 0994 0994C4 bad posture 3 6000 0999 0999 0999
0997 0997 0997
will develop a software application that suggests correctinghisher bad posture to a right posture during desk work
43 Indoor Localization by Sensorized Glasses We have pro-posed an indoor localization system using illuminance-basedtrilateration method [22] In this system we assume that theperson wears glasses having an illuminance sensor as shownin Figure 13 and controllable light sources are deployed in thetarget environment Our system controls lights andmeasuresthe variation of illuminance at the glasses As shown inFigure 14 the distances (1198891ndash1198893) from each light (1198971ndash1198973) to theuser (119906) are estimated from the distance-illumination modelconstructed in advance and userrsquos position is estimated basedon trilateration method using estimated distances
As a result of an experiment where three controllablelight sources were deployed in a room we confirmed thatour system could estimate the position within 05m erroron average By using SenStick we could easily develop thesensorized glasses equipped with illuminance sensor
44 Sensorized Airsoft Gun We have developed the sen-sorized airsoft gun for activity recognition in airsoft gameas shown in Figure 15 Airsoft game is a sport in whichparticipants eliminate opponents by hitting each other usingreplica weapons called airsoft guns The graph of a typicalsignal from the sensorized airsoft gun is shown in Figure 16A user can count the number of shots of a ball by attachingSenStick to his airsoft gun
45 Another Example Studentrsquos Unique Achievements andSurvey Results SenStick is also used in a lecture of IoT
Table 8 Confusion matrix of posture type recognition
PredictedC1 C2 C3 C4
ActualC1 good posture 1000 0000 0000 0000C2 bad posture 1 0000 0994 0006 0001C3 bad posture 2 0000 0006 0994 0000C4 bad posture 3 0000 0001 0000 0999
prototyping Students realize their unique IoT devices easilyby using SenSticks Here we introduce some of the developedIoT devices and survey results about the usefulness ofSenStick answered by students
Figure 17 shows a sensorized pencil with two SenSticksThe sensorized pencil is developed to recognize the charac-ters written by this pencil Even though two SenSticks areembedded to detect precise movement of pencil the size ofthe pencil is reasonable
Figure 18 shows a sensorized name tag anddigital signageSince we changed the battery from 60mAh to 900mAh thetag keeps running more than one week It aims to recordemployeesrsquo or studentsrsquo daily activities to support their healthThe digital signage informs the user who wears a sensorizedname tag about the userrsquos activity data when the user standsin front of it
Figure 19 shows a sensorized spoon used in the digitalfish catching game for children In this game the sensorizedspoon detects the scooping movement of the hand Thuschildren can catch the digital fish displayed by the projectionmapping by using sensorized spoonThis systemwas actuallyused at the universityrsquos open campus event and around100 children enjoyed this game and showed interest in IoTtechnology
The survey results obtained from the ten graduate stu-dents (ninemales and one female) who developed prototypesas mentioned above are shown in Table 9 A questionnaireis a five-grade form (5 strongly agree 4 agree a little 3neither agree nor disagree 2 disagree a little and 1 stronglydisagree) From the results of questions 1 and 2 manyparticipants think that SenStick platform is useful for IoT
14 Journal of Sensors
Position
Distance illumination model
Using trilateration method
a
Light 3
Light 2
Light 1
L = minus421d + 1496
L = minus063d + 501
05 1 15 2 25 3 35 4 450Distance between the illuminance sensor and the lighting device (m)
02468
1012141618
Illum
inan
ce v
alue
(lux
)
Distanceu
d1
d2
d3
l1 p1 x1
l3 p3 x3
l2 p2 x2
Figure 14 Indoor localization by sensorized glasses
Sensorized airsoft gun
Figure 15 Sensorized airsoft gun
research and IoT education Also the results of question 3show that many students are interested in the SenStick andwant to use it again From the results of questions 4 and 5we can see that usability and specification of SenStick arerelatively high There were opinions about the usefulness ofSenStick platform such as ldquoI could quickly try out ideas of IoTprototypingrdquo ldquoIt was easy to collect sensing datardquo and ldquoSinceSenStick is small it was easily attached tomost of the thingsrdquoOn the other hand there were opinions about improvements
such as ldquoI want to use higher sensing frequency (50ndash100Hz)rdquoldquoI want to process a sensing data on SenStickrdquo and ldquoI want tomodify the firmware itself depending on the application ofthe IoT devicerdquo
5 Conclusion
In this paper we have introduced a smart and comprehensivesensing platform called SenStick which is composed of
Journal of Sensors 15
Typical signal of airso gun
Hold a gun Ready
Shoot
1 2 3 4 5 6 70
00
30
60
91
21
51
82
12
42
73
03
33
63
94
24
54
85
15
45
76
06
36
66
97
27
57
88
18
48
79
09
39
69
910
210
510
811
111
411
712
012
312
612
913
2
minus600
minus400
minus200
0
200
400
600
800
acc_xacc_yacc_z
gyro_xgyro_y
acc_synthesis gyro_zgyro_synthesis
Figure 16 Time series of sensing data in shooting an airsoft gun
Sensorized pencil
Figure 17 Sensorized pencil
Sensorized name tagVisualization of sensor data
Figure 18 Sensorized name tag
Sensorized spoon Digital fishing game
Figure 19 Sensorized spoon
Table 9 Questionnaire results about SenStick
Question contents ScoreQ1 Do you think that SenStick will be useful for IoTresearch 49
Q2 Do you think that SenStick will be useful for IoTeducation 47
Q3 Do you want to use SenStick again 49Q4 Do you think that SenStick was easy to use 41Q5 Are you satisfied with specification of SenStick 41
hardware software and 3D case data The main purpose ofour platform is to simplify the sensorization of our personalbelongings
The hardware of SenStick is maybe one of the worldrsquossmallest sensor boards which includes 8 kinds of sensorsand BLE In addition to the hardware we also developedsupport software for managing monitoring recording andviewing the hardware By providing 3D case data all theresearchers can measure every tiny activity in daily life easilyand efficiently For an educational use it is also useful forunderstanding sensors and prototyping the desired systemincluding tiny censored things
Conflicts of Interest
The authors declare that there are no conflicts of interestregarding the publication of this paper
Acknowledgments
This work is supported in part by JST PRESTO and JSPSKAKENHI (Grant no 17J10021) and ldquoResearch and Devel-opment of Innovative Network Technologies to Create theFuturerdquo the Commissioned Research of National Instituteof Information and Communications Technology (NICT)Japan
16 Journal of Sensors
References
[1] O Brdiczka M Langet J Maisonnasse and J L CrowleyldquoDetecting human behavior models from multimodal observa-tion in a smart homerdquo IEEE Transactions on Automation Scienceand Engineering vol 6 no 4 pp 588ndash597 2009
[2] J Chen A Kam J Zhang N Liu and L Shue ldquoBathroomactivity monitoring based on soundrdquo in Pervasive ComputingThird International Conference PERVASIVE 2005 MunichGermanyMay 8-13 2005 Proceedings vol 3468 ofLectureNotesin Computer Science pp 47ndash61 Springer Berlin Germany2002
[3] G Bauer K Stockinger and P Lukowicz ldquoRecognizing theUse-Mode of Kitchen Appliances from Their Current Consump-tionrdquo in Smart Sensing and Context vol 5741 of Lecture Notesin Computer Science pp 163ndash176 Springer Berlin HeidelbergBerlin Heidelberg 2009
[4] K Ouchi and M Doi ldquoIndoor-outdoor activity recognition bya smartphonerdquo in Proceedings of the the 2012 ACM Conferencep 537 Pittsburgh Pennsylvania September 2012
[5] A Anjum and M U Ilyas ldquoActivity recognition using smart-phone sensorsrdquo in Proceedings of the IEEE 10th ConsumerCommunications and Networking Conference (CCNC rsquo13) pp914ndash919 IEEE January 2013
[6] O D Lara and M A Labrador ldquoA survey on human activityrecognition using wearable sensorsrdquo IEEE CommunicationsSurveys amp Tutorials vol 15 no 3 pp 1192ndash1209 2013
[7] ldquoJins memerdquo httpsjins-memecom[8] Y Matsuda T Ishioka T Hasegawa et al ldquoWaistonBelt A belt
for monitoring your real abdominal circumference foreverrdquo inProceedings of the ACM International Joint Conference on Perva-sive and Ubiquitous Computing and the 2015 ACM InternationalSymposium on Wearable Computers UbiComp and ISWC 2015pp 329ndash332 September 2015
[9] M Kurahashi K Murao T Terada and M Tsukamoto ldquoPer-sonal identification system based on rotation of toilet paperrollsrdquo in Proceedings of the IEEE International Conference onPervasive Computing and CommunicationsWorkshops (PerComWorkshops rsquo17) pp 521ndash526 IEEE Kona Hawaii USA March2017
[10] ldquoHapiforkrdquo httpswwwhapicom[11] Y Arakawa ldquoSenStick Sensorize every thingsrdquo in Proceedings of
the ACM International Joint Conference on Pervasive and Ubiq-uitous Computing and the 2015 ACM International SymposiumonWearable Computers UbiComp and ISWC 2015 pp 349ndash352September 2015
[12] Y Nakamura T Kanehira Y Arakawa and K YasumotoldquoSenStick 2 Ultra tiny all-in-one sensor with wireless chargingrdquoinProceedings of the 2016ACM International Joint Conference onPervasive and Ubiquitous Computing UbiComp 2016 pp 337ndash340 September 2016
[13] ldquoSensortag cc2541 (texas instruments)rdquo httpwwwticomtoolcc2541dk-sensor
[14] ldquoIot smart modulerdquo httpwwwalpscomjiotsmart[15] ldquoTsnd121 (atr-promotions)rdquo httpwwwatr-pcomproducts
TSND121html[16] ldquoWax9 (axivity)rdquo httpaxivitycomproductwax9[17] Uk biobank ldquoUk biobankrdquo httpwwwukbiobankacuk[18] J C Seidell and J Halberstadt ldquoThe global burden of obesity
and the challenges of preventionrdquo Annals of Nutrition andMetabolism vol 66 supplement 2 pp 7ndash12 2015
[19] ldquoWeltThe smart belt for fashion and healthrdquo httpswwwwelt-corpcom
[20] ldquoBelty power The belt your really deserverdquo httpswwwbeltyco
[21] H Nam J-H Kim and J-I Kim ldquoSmart Belt A wearabledevice for managing abdominal obesityrdquo in Proceedings of theInternational Conference on Big Data and Smart ComputingBigComp 2016 pp 430ndash434 January 2016
[22] K Moriya M Fujimoto Y Arakawa and K Yasumoto ldquoIndoorlocalization based on distance-illuminance model and activecontrol of lighting devicesrdquo in Proceedings of the 2016 Interna-tional Conference on Indoor Positioning and Indoor NavigationIPIN 2016 pp 1ndash6 October 2016
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal of
Volume 201
Submit your manuscripts athttpswwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
12 Journal of Sensors
Enter the lab
HumidityTemp
Humidity dropping
Pressure dropping
Working
Walking WalkingClimbing stairs Sitting (good posture) Sitting (bad posture)
AccsXAccsYAccsZ
60
40
20
1003
1002
1001
1000
999
25
2
15
1
05
0
minus15
minus1
minus05
Air pressure
Figure 11 Time series of sensing data in daily activity
C1 good posture C2 bad posture 1
C3 bad posture 2 C4 bad posture 3
Figure 12 Time series of sensing data in posture monitoring
angle of the belt calculated from 3 axial acceleration sensorsWe set the sensing rate as 10Hz and collected 6000 samples(10min) for each posture We assumed 4 classes (C1 goodposture C2 bad posture 1 C3 bad posture 2 and C4 badposture 3) as shown in Figure 12
In this experiment we calculated the recognition accu-racy of posture type classification by using Random Forestclassifier of Weka Tables 7 and 8 show the classificationresults The average recognition accuracy was 997 Inthe future based on the result of posture estimation we
Journal of Sensors 13
Sensorized glasses
Figure 13 Sensorized glasses
Table 7 Experimental result (posture type)
Posture type Number ofpieces of data Precision Recall 119865-Measure
C1 good posture 6000 1000 1000 1000C2 bad posture 1 6000 0994 0994 0994C3 bad posture 2 6000 0994 0994 0994C4 bad posture 3 6000 0999 0999 0999
0997 0997 0997
will develop a software application that suggests correctinghisher bad posture to a right posture during desk work
43 Indoor Localization by Sensorized Glasses We have pro-posed an indoor localization system using illuminance-basedtrilateration method [22] In this system we assume that theperson wears glasses having an illuminance sensor as shownin Figure 13 and controllable light sources are deployed in thetarget environment Our system controls lights andmeasuresthe variation of illuminance at the glasses As shown inFigure 14 the distances (1198891ndash1198893) from each light (1198971ndash1198973) to theuser (119906) are estimated from the distance-illumination modelconstructed in advance and userrsquos position is estimated basedon trilateration method using estimated distances
As a result of an experiment where three controllablelight sources were deployed in a room we confirmed thatour system could estimate the position within 05m erroron average By using SenStick we could easily develop thesensorized glasses equipped with illuminance sensor
44 Sensorized Airsoft Gun We have developed the sen-sorized airsoft gun for activity recognition in airsoft gameas shown in Figure 15 Airsoft game is a sport in whichparticipants eliminate opponents by hitting each other usingreplica weapons called airsoft guns The graph of a typicalsignal from the sensorized airsoft gun is shown in Figure 16A user can count the number of shots of a ball by attachingSenStick to his airsoft gun
45 Another Example Studentrsquos Unique Achievements andSurvey Results SenStick is also used in a lecture of IoT
Table 8 Confusion matrix of posture type recognition
PredictedC1 C2 C3 C4
ActualC1 good posture 1000 0000 0000 0000C2 bad posture 1 0000 0994 0006 0001C3 bad posture 2 0000 0006 0994 0000C4 bad posture 3 0000 0001 0000 0999
prototyping Students realize their unique IoT devices easilyby using SenSticks Here we introduce some of the developedIoT devices and survey results about the usefulness ofSenStick answered by students
Figure 17 shows a sensorized pencil with two SenSticksThe sensorized pencil is developed to recognize the charac-ters written by this pencil Even though two SenSticks areembedded to detect precise movement of pencil the size ofthe pencil is reasonable
Figure 18 shows a sensorized name tag anddigital signageSince we changed the battery from 60mAh to 900mAh thetag keeps running more than one week It aims to recordemployeesrsquo or studentsrsquo daily activities to support their healthThe digital signage informs the user who wears a sensorizedname tag about the userrsquos activity data when the user standsin front of it
Figure 19 shows a sensorized spoon used in the digitalfish catching game for children In this game the sensorizedspoon detects the scooping movement of the hand Thuschildren can catch the digital fish displayed by the projectionmapping by using sensorized spoonThis systemwas actuallyused at the universityrsquos open campus event and around100 children enjoyed this game and showed interest in IoTtechnology
The survey results obtained from the ten graduate stu-dents (ninemales and one female) who developed prototypesas mentioned above are shown in Table 9 A questionnaireis a five-grade form (5 strongly agree 4 agree a little 3neither agree nor disagree 2 disagree a little and 1 stronglydisagree) From the results of questions 1 and 2 manyparticipants think that SenStick platform is useful for IoT
14 Journal of Sensors
Position
Distance illumination model
Using trilateration method
a
Light 3
Light 2
Light 1
L = minus421d + 1496
L = minus063d + 501
05 1 15 2 25 3 35 4 450Distance between the illuminance sensor and the lighting device (m)
02468
1012141618
Illum
inan
ce v
alue
(lux
)
Distanceu
d1
d2
d3
l1 p1 x1
l3 p3 x3
l2 p2 x2
Figure 14 Indoor localization by sensorized glasses
Sensorized airsoft gun
Figure 15 Sensorized airsoft gun
research and IoT education Also the results of question 3show that many students are interested in the SenStick andwant to use it again From the results of questions 4 and 5we can see that usability and specification of SenStick arerelatively high There were opinions about the usefulness ofSenStick platform such as ldquoI could quickly try out ideas of IoTprototypingrdquo ldquoIt was easy to collect sensing datardquo and ldquoSinceSenStick is small it was easily attached tomost of the thingsrdquoOn the other hand there were opinions about improvements
such as ldquoI want to use higher sensing frequency (50ndash100Hz)rdquoldquoI want to process a sensing data on SenStickrdquo and ldquoI want tomodify the firmware itself depending on the application ofthe IoT devicerdquo
5 Conclusion
In this paper we have introduced a smart and comprehensivesensing platform called SenStick which is composed of
Journal of Sensors 15
Typical signal of airso gun
Hold a gun Ready
Shoot
1 2 3 4 5 6 70
00
30
60
91
21
51
82
12
42
73
03
33
63
94
24
54
85
15
45
76
06
36
66
97
27
57
88
18
48
79
09
39
69
910
210
510
811
111
411
712
012
312
612
913
2
minus600
minus400
minus200
0
200
400
600
800
acc_xacc_yacc_z
gyro_xgyro_y
acc_synthesis gyro_zgyro_synthesis
Figure 16 Time series of sensing data in shooting an airsoft gun
Sensorized pencil
Figure 17 Sensorized pencil
Sensorized name tagVisualization of sensor data
Figure 18 Sensorized name tag
Sensorized spoon Digital fishing game
Figure 19 Sensorized spoon
Table 9 Questionnaire results about SenStick
Question contents ScoreQ1 Do you think that SenStick will be useful for IoTresearch 49
Q2 Do you think that SenStick will be useful for IoTeducation 47
Q3 Do you want to use SenStick again 49Q4 Do you think that SenStick was easy to use 41Q5 Are you satisfied with specification of SenStick 41
hardware software and 3D case data The main purpose ofour platform is to simplify the sensorization of our personalbelongings
The hardware of SenStick is maybe one of the worldrsquossmallest sensor boards which includes 8 kinds of sensorsand BLE In addition to the hardware we also developedsupport software for managing monitoring recording andviewing the hardware By providing 3D case data all theresearchers can measure every tiny activity in daily life easilyand efficiently For an educational use it is also useful forunderstanding sensors and prototyping the desired systemincluding tiny censored things
Conflicts of Interest
The authors declare that there are no conflicts of interestregarding the publication of this paper
Acknowledgments
This work is supported in part by JST PRESTO and JSPSKAKENHI (Grant no 17J10021) and ldquoResearch and Devel-opment of Innovative Network Technologies to Create theFuturerdquo the Commissioned Research of National Instituteof Information and Communications Technology (NICT)Japan
16 Journal of Sensors
References
[1] O Brdiczka M Langet J Maisonnasse and J L CrowleyldquoDetecting human behavior models from multimodal observa-tion in a smart homerdquo IEEE Transactions on Automation Scienceand Engineering vol 6 no 4 pp 588ndash597 2009
[2] J Chen A Kam J Zhang N Liu and L Shue ldquoBathroomactivity monitoring based on soundrdquo in Pervasive ComputingThird International Conference PERVASIVE 2005 MunichGermanyMay 8-13 2005 Proceedings vol 3468 ofLectureNotesin Computer Science pp 47ndash61 Springer Berlin Germany2002
[3] G Bauer K Stockinger and P Lukowicz ldquoRecognizing theUse-Mode of Kitchen Appliances from Their Current Consump-tionrdquo in Smart Sensing and Context vol 5741 of Lecture Notesin Computer Science pp 163ndash176 Springer Berlin HeidelbergBerlin Heidelberg 2009
[4] K Ouchi and M Doi ldquoIndoor-outdoor activity recognition bya smartphonerdquo in Proceedings of the the 2012 ACM Conferencep 537 Pittsburgh Pennsylvania September 2012
[5] A Anjum and M U Ilyas ldquoActivity recognition using smart-phone sensorsrdquo in Proceedings of the IEEE 10th ConsumerCommunications and Networking Conference (CCNC rsquo13) pp914ndash919 IEEE January 2013
[6] O D Lara and M A Labrador ldquoA survey on human activityrecognition using wearable sensorsrdquo IEEE CommunicationsSurveys amp Tutorials vol 15 no 3 pp 1192ndash1209 2013
[7] ldquoJins memerdquo httpsjins-memecom[8] Y Matsuda T Ishioka T Hasegawa et al ldquoWaistonBelt A belt
for monitoring your real abdominal circumference foreverrdquo inProceedings of the ACM International Joint Conference on Perva-sive and Ubiquitous Computing and the 2015 ACM InternationalSymposium on Wearable Computers UbiComp and ISWC 2015pp 329ndash332 September 2015
[9] M Kurahashi K Murao T Terada and M Tsukamoto ldquoPer-sonal identification system based on rotation of toilet paperrollsrdquo in Proceedings of the IEEE International Conference onPervasive Computing and CommunicationsWorkshops (PerComWorkshops rsquo17) pp 521ndash526 IEEE Kona Hawaii USA March2017
[10] ldquoHapiforkrdquo httpswwwhapicom[11] Y Arakawa ldquoSenStick Sensorize every thingsrdquo in Proceedings of
the ACM International Joint Conference on Pervasive and Ubiq-uitous Computing and the 2015 ACM International SymposiumonWearable Computers UbiComp and ISWC 2015 pp 349ndash352September 2015
[12] Y Nakamura T Kanehira Y Arakawa and K YasumotoldquoSenStick 2 Ultra tiny all-in-one sensor with wireless chargingrdquoinProceedings of the 2016ACM International Joint Conference onPervasive and Ubiquitous Computing UbiComp 2016 pp 337ndash340 September 2016
[13] ldquoSensortag cc2541 (texas instruments)rdquo httpwwwticomtoolcc2541dk-sensor
[14] ldquoIot smart modulerdquo httpwwwalpscomjiotsmart[15] ldquoTsnd121 (atr-promotions)rdquo httpwwwatr-pcomproducts
TSND121html[16] ldquoWax9 (axivity)rdquo httpaxivitycomproductwax9[17] Uk biobank ldquoUk biobankrdquo httpwwwukbiobankacuk[18] J C Seidell and J Halberstadt ldquoThe global burden of obesity
and the challenges of preventionrdquo Annals of Nutrition andMetabolism vol 66 supplement 2 pp 7ndash12 2015
[19] ldquoWeltThe smart belt for fashion and healthrdquo httpswwwwelt-corpcom
[20] ldquoBelty power The belt your really deserverdquo httpswwwbeltyco
[21] H Nam J-H Kim and J-I Kim ldquoSmart Belt A wearabledevice for managing abdominal obesityrdquo in Proceedings of theInternational Conference on Big Data and Smart ComputingBigComp 2016 pp 430ndash434 January 2016
[22] K Moriya M Fujimoto Y Arakawa and K Yasumoto ldquoIndoorlocalization based on distance-illuminance model and activecontrol of lighting devicesrdquo in Proceedings of the 2016 Interna-tional Conference on Indoor Positioning and Indoor NavigationIPIN 2016 pp 1ndash6 October 2016
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal of
Volume 201
Submit your manuscripts athttpswwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
Journal of Sensors 13
Sensorized glasses
Figure 13 Sensorized glasses
Table 7 Experimental result (posture type)
Posture type Number ofpieces of data Precision Recall 119865-Measure
C1 good posture 6000 1000 1000 1000C2 bad posture 1 6000 0994 0994 0994C3 bad posture 2 6000 0994 0994 0994C4 bad posture 3 6000 0999 0999 0999
0997 0997 0997
will develop a software application that suggests correctinghisher bad posture to a right posture during desk work
43 Indoor Localization by Sensorized Glasses We have pro-posed an indoor localization system using illuminance-basedtrilateration method [22] In this system we assume that theperson wears glasses having an illuminance sensor as shownin Figure 13 and controllable light sources are deployed in thetarget environment Our system controls lights andmeasuresthe variation of illuminance at the glasses As shown inFigure 14 the distances (1198891ndash1198893) from each light (1198971ndash1198973) to theuser (119906) are estimated from the distance-illumination modelconstructed in advance and userrsquos position is estimated basedon trilateration method using estimated distances
As a result of an experiment where three controllablelight sources were deployed in a room we confirmed thatour system could estimate the position within 05m erroron average By using SenStick we could easily develop thesensorized glasses equipped with illuminance sensor
44 Sensorized Airsoft Gun We have developed the sen-sorized airsoft gun for activity recognition in airsoft gameas shown in Figure 15 Airsoft game is a sport in whichparticipants eliminate opponents by hitting each other usingreplica weapons called airsoft guns The graph of a typicalsignal from the sensorized airsoft gun is shown in Figure 16A user can count the number of shots of a ball by attachingSenStick to his airsoft gun
45 Another Example Studentrsquos Unique Achievements andSurvey Results SenStick is also used in a lecture of IoT
Table 8 Confusion matrix of posture type recognition
PredictedC1 C2 C3 C4
ActualC1 good posture 1000 0000 0000 0000C2 bad posture 1 0000 0994 0006 0001C3 bad posture 2 0000 0006 0994 0000C4 bad posture 3 0000 0001 0000 0999
prototyping Students realize their unique IoT devices easilyby using SenSticks Here we introduce some of the developedIoT devices and survey results about the usefulness ofSenStick answered by students
Figure 17 shows a sensorized pencil with two SenSticksThe sensorized pencil is developed to recognize the charac-ters written by this pencil Even though two SenSticks areembedded to detect precise movement of pencil the size ofthe pencil is reasonable
Figure 18 shows a sensorized name tag anddigital signageSince we changed the battery from 60mAh to 900mAh thetag keeps running more than one week It aims to recordemployeesrsquo or studentsrsquo daily activities to support their healthThe digital signage informs the user who wears a sensorizedname tag about the userrsquos activity data when the user standsin front of it
Figure 19 shows a sensorized spoon used in the digitalfish catching game for children In this game the sensorizedspoon detects the scooping movement of the hand Thuschildren can catch the digital fish displayed by the projectionmapping by using sensorized spoonThis systemwas actuallyused at the universityrsquos open campus event and around100 children enjoyed this game and showed interest in IoTtechnology
The survey results obtained from the ten graduate stu-dents (ninemales and one female) who developed prototypesas mentioned above are shown in Table 9 A questionnaireis a five-grade form (5 strongly agree 4 agree a little 3neither agree nor disagree 2 disagree a little and 1 stronglydisagree) From the results of questions 1 and 2 manyparticipants think that SenStick platform is useful for IoT
14 Journal of Sensors
Position
Distance illumination model
Using trilateration method
a
Light 3
Light 2
Light 1
L = minus421d + 1496
L = minus063d + 501
05 1 15 2 25 3 35 4 450Distance between the illuminance sensor and the lighting device (m)
02468
1012141618
Illum
inan
ce v
alue
(lux
)
Distanceu
d1
d2
d3
l1 p1 x1
l3 p3 x3
l2 p2 x2
Figure 14 Indoor localization by sensorized glasses
Sensorized airsoft gun
Figure 15 Sensorized airsoft gun
research and IoT education Also the results of question 3show that many students are interested in the SenStick andwant to use it again From the results of questions 4 and 5we can see that usability and specification of SenStick arerelatively high There were opinions about the usefulness ofSenStick platform such as ldquoI could quickly try out ideas of IoTprototypingrdquo ldquoIt was easy to collect sensing datardquo and ldquoSinceSenStick is small it was easily attached tomost of the thingsrdquoOn the other hand there were opinions about improvements
such as ldquoI want to use higher sensing frequency (50ndash100Hz)rdquoldquoI want to process a sensing data on SenStickrdquo and ldquoI want tomodify the firmware itself depending on the application ofthe IoT devicerdquo
5 Conclusion
In this paper we have introduced a smart and comprehensivesensing platform called SenStick which is composed of
Journal of Sensors 15
Typical signal of airso gun
Hold a gun Ready
Shoot
1 2 3 4 5 6 70
00
30
60
91
21
51
82
12
42
73
03
33
63
94
24
54
85
15
45
76
06
36
66
97
27
57
88
18
48
79
09
39
69
910
210
510
811
111
411
712
012
312
612
913
2
minus600
minus400
minus200
0
200
400
600
800
acc_xacc_yacc_z
gyro_xgyro_y
acc_synthesis gyro_zgyro_synthesis
Figure 16 Time series of sensing data in shooting an airsoft gun
Sensorized pencil
Figure 17 Sensorized pencil
Sensorized name tagVisualization of sensor data
Figure 18 Sensorized name tag
Sensorized spoon Digital fishing game
Figure 19 Sensorized spoon
Table 9 Questionnaire results about SenStick
Question contents ScoreQ1 Do you think that SenStick will be useful for IoTresearch 49
Q2 Do you think that SenStick will be useful for IoTeducation 47
Q3 Do you want to use SenStick again 49Q4 Do you think that SenStick was easy to use 41Q5 Are you satisfied with specification of SenStick 41
hardware software and 3D case data The main purpose ofour platform is to simplify the sensorization of our personalbelongings
The hardware of SenStick is maybe one of the worldrsquossmallest sensor boards which includes 8 kinds of sensorsand BLE In addition to the hardware we also developedsupport software for managing monitoring recording andviewing the hardware By providing 3D case data all theresearchers can measure every tiny activity in daily life easilyand efficiently For an educational use it is also useful forunderstanding sensors and prototyping the desired systemincluding tiny censored things
Conflicts of Interest
The authors declare that there are no conflicts of interestregarding the publication of this paper
Acknowledgments
This work is supported in part by JST PRESTO and JSPSKAKENHI (Grant no 17J10021) and ldquoResearch and Devel-opment of Innovative Network Technologies to Create theFuturerdquo the Commissioned Research of National Instituteof Information and Communications Technology (NICT)Japan
16 Journal of Sensors
References
[1] O Brdiczka M Langet J Maisonnasse and J L CrowleyldquoDetecting human behavior models from multimodal observa-tion in a smart homerdquo IEEE Transactions on Automation Scienceand Engineering vol 6 no 4 pp 588ndash597 2009
[2] J Chen A Kam J Zhang N Liu and L Shue ldquoBathroomactivity monitoring based on soundrdquo in Pervasive ComputingThird International Conference PERVASIVE 2005 MunichGermanyMay 8-13 2005 Proceedings vol 3468 ofLectureNotesin Computer Science pp 47ndash61 Springer Berlin Germany2002
[3] G Bauer K Stockinger and P Lukowicz ldquoRecognizing theUse-Mode of Kitchen Appliances from Their Current Consump-tionrdquo in Smart Sensing and Context vol 5741 of Lecture Notesin Computer Science pp 163ndash176 Springer Berlin HeidelbergBerlin Heidelberg 2009
[4] K Ouchi and M Doi ldquoIndoor-outdoor activity recognition bya smartphonerdquo in Proceedings of the the 2012 ACM Conferencep 537 Pittsburgh Pennsylvania September 2012
[5] A Anjum and M U Ilyas ldquoActivity recognition using smart-phone sensorsrdquo in Proceedings of the IEEE 10th ConsumerCommunications and Networking Conference (CCNC rsquo13) pp914ndash919 IEEE January 2013
[6] O D Lara and M A Labrador ldquoA survey on human activityrecognition using wearable sensorsrdquo IEEE CommunicationsSurveys amp Tutorials vol 15 no 3 pp 1192ndash1209 2013
[7] ldquoJins memerdquo httpsjins-memecom[8] Y Matsuda T Ishioka T Hasegawa et al ldquoWaistonBelt A belt
for monitoring your real abdominal circumference foreverrdquo inProceedings of the ACM International Joint Conference on Perva-sive and Ubiquitous Computing and the 2015 ACM InternationalSymposium on Wearable Computers UbiComp and ISWC 2015pp 329ndash332 September 2015
[9] M Kurahashi K Murao T Terada and M Tsukamoto ldquoPer-sonal identification system based on rotation of toilet paperrollsrdquo in Proceedings of the IEEE International Conference onPervasive Computing and CommunicationsWorkshops (PerComWorkshops rsquo17) pp 521ndash526 IEEE Kona Hawaii USA March2017
[10] ldquoHapiforkrdquo httpswwwhapicom[11] Y Arakawa ldquoSenStick Sensorize every thingsrdquo in Proceedings of
the ACM International Joint Conference on Pervasive and Ubiq-uitous Computing and the 2015 ACM International SymposiumonWearable Computers UbiComp and ISWC 2015 pp 349ndash352September 2015
[12] Y Nakamura T Kanehira Y Arakawa and K YasumotoldquoSenStick 2 Ultra tiny all-in-one sensor with wireless chargingrdquoinProceedings of the 2016ACM International Joint Conference onPervasive and Ubiquitous Computing UbiComp 2016 pp 337ndash340 September 2016
[13] ldquoSensortag cc2541 (texas instruments)rdquo httpwwwticomtoolcc2541dk-sensor
[14] ldquoIot smart modulerdquo httpwwwalpscomjiotsmart[15] ldquoTsnd121 (atr-promotions)rdquo httpwwwatr-pcomproducts
TSND121html[16] ldquoWax9 (axivity)rdquo httpaxivitycomproductwax9[17] Uk biobank ldquoUk biobankrdquo httpwwwukbiobankacuk[18] J C Seidell and J Halberstadt ldquoThe global burden of obesity
and the challenges of preventionrdquo Annals of Nutrition andMetabolism vol 66 supplement 2 pp 7ndash12 2015
[19] ldquoWeltThe smart belt for fashion and healthrdquo httpswwwwelt-corpcom
[20] ldquoBelty power The belt your really deserverdquo httpswwwbeltyco
[21] H Nam J-H Kim and J-I Kim ldquoSmart Belt A wearabledevice for managing abdominal obesityrdquo in Proceedings of theInternational Conference on Big Data and Smart ComputingBigComp 2016 pp 430ndash434 January 2016
[22] K Moriya M Fujimoto Y Arakawa and K Yasumoto ldquoIndoorlocalization based on distance-illuminance model and activecontrol of lighting devicesrdquo in Proceedings of the 2016 Interna-tional Conference on Indoor Positioning and Indoor NavigationIPIN 2016 pp 1ndash6 October 2016
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal of
Volume 201
Submit your manuscripts athttpswwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
14 Journal of Sensors
Position
Distance illumination model
Using trilateration method
a
Light 3
Light 2
Light 1
L = minus421d + 1496
L = minus063d + 501
05 1 15 2 25 3 35 4 450Distance between the illuminance sensor and the lighting device (m)
02468
1012141618
Illum
inan
ce v
alue
(lux
)
Distanceu
d1
d2
d3
l1 p1 x1
l3 p3 x3
l2 p2 x2
Figure 14 Indoor localization by sensorized glasses
Sensorized airsoft gun
Figure 15 Sensorized airsoft gun
research and IoT education Also the results of question 3show that many students are interested in the SenStick andwant to use it again From the results of questions 4 and 5we can see that usability and specification of SenStick arerelatively high There were opinions about the usefulness ofSenStick platform such as ldquoI could quickly try out ideas of IoTprototypingrdquo ldquoIt was easy to collect sensing datardquo and ldquoSinceSenStick is small it was easily attached tomost of the thingsrdquoOn the other hand there were opinions about improvements
such as ldquoI want to use higher sensing frequency (50ndash100Hz)rdquoldquoI want to process a sensing data on SenStickrdquo and ldquoI want tomodify the firmware itself depending on the application ofthe IoT devicerdquo
5 Conclusion
In this paper we have introduced a smart and comprehensivesensing platform called SenStick which is composed of
Journal of Sensors 15
Typical signal of airso gun
Hold a gun Ready
Shoot
1 2 3 4 5 6 70
00
30
60
91
21
51
82
12
42
73
03
33
63
94
24
54
85
15
45
76
06
36
66
97
27
57
88
18
48
79
09
39
69
910
210
510
811
111
411
712
012
312
612
913
2
minus600
minus400
minus200
0
200
400
600
800
acc_xacc_yacc_z
gyro_xgyro_y
acc_synthesis gyro_zgyro_synthesis
Figure 16 Time series of sensing data in shooting an airsoft gun
Sensorized pencil
Figure 17 Sensorized pencil
Sensorized name tagVisualization of sensor data
Figure 18 Sensorized name tag
Sensorized spoon Digital fishing game
Figure 19 Sensorized spoon
Table 9 Questionnaire results about SenStick
Question contents ScoreQ1 Do you think that SenStick will be useful for IoTresearch 49
Q2 Do you think that SenStick will be useful for IoTeducation 47
Q3 Do you want to use SenStick again 49Q4 Do you think that SenStick was easy to use 41Q5 Are you satisfied with specification of SenStick 41
hardware software and 3D case data The main purpose ofour platform is to simplify the sensorization of our personalbelongings
The hardware of SenStick is maybe one of the worldrsquossmallest sensor boards which includes 8 kinds of sensorsand BLE In addition to the hardware we also developedsupport software for managing monitoring recording andviewing the hardware By providing 3D case data all theresearchers can measure every tiny activity in daily life easilyand efficiently For an educational use it is also useful forunderstanding sensors and prototyping the desired systemincluding tiny censored things
Conflicts of Interest
The authors declare that there are no conflicts of interestregarding the publication of this paper
Acknowledgments
This work is supported in part by JST PRESTO and JSPSKAKENHI (Grant no 17J10021) and ldquoResearch and Devel-opment of Innovative Network Technologies to Create theFuturerdquo the Commissioned Research of National Instituteof Information and Communications Technology (NICT)Japan
16 Journal of Sensors
References
[1] O Brdiczka M Langet J Maisonnasse and J L CrowleyldquoDetecting human behavior models from multimodal observa-tion in a smart homerdquo IEEE Transactions on Automation Scienceand Engineering vol 6 no 4 pp 588ndash597 2009
[2] J Chen A Kam J Zhang N Liu and L Shue ldquoBathroomactivity monitoring based on soundrdquo in Pervasive ComputingThird International Conference PERVASIVE 2005 MunichGermanyMay 8-13 2005 Proceedings vol 3468 ofLectureNotesin Computer Science pp 47ndash61 Springer Berlin Germany2002
[3] G Bauer K Stockinger and P Lukowicz ldquoRecognizing theUse-Mode of Kitchen Appliances from Their Current Consump-tionrdquo in Smart Sensing and Context vol 5741 of Lecture Notesin Computer Science pp 163ndash176 Springer Berlin HeidelbergBerlin Heidelberg 2009
[4] K Ouchi and M Doi ldquoIndoor-outdoor activity recognition bya smartphonerdquo in Proceedings of the the 2012 ACM Conferencep 537 Pittsburgh Pennsylvania September 2012
[5] A Anjum and M U Ilyas ldquoActivity recognition using smart-phone sensorsrdquo in Proceedings of the IEEE 10th ConsumerCommunications and Networking Conference (CCNC rsquo13) pp914ndash919 IEEE January 2013
[6] O D Lara and M A Labrador ldquoA survey on human activityrecognition using wearable sensorsrdquo IEEE CommunicationsSurveys amp Tutorials vol 15 no 3 pp 1192ndash1209 2013
[7] ldquoJins memerdquo httpsjins-memecom[8] Y Matsuda T Ishioka T Hasegawa et al ldquoWaistonBelt A belt
for monitoring your real abdominal circumference foreverrdquo inProceedings of the ACM International Joint Conference on Perva-sive and Ubiquitous Computing and the 2015 ACM InternationalSymposium on Wearable Computers UbiComp and ISWC 2015pp 329ndash332 September 2015
[9] M Kurahashi K Murao T Terada and M Tsukamoto ldquoPer-sonal identification system based on rotation of toilet paperrollsrdquo in Proceedings of the IEEE International Conference onPervasive Computing and CommunicationsWorkshops (PerComWorkshops rsquo17) pp 521ndash526 IEEE Kona Hawaii USA March2017
[10] ldquoHapiforkrdquo httpswwwhapicom[11] Y Arakawa ldquoSenStick Sensorize every thingsrdquo in Proceedings of
the ACM International Joint Conference on Pervasive and Ubiq-uitous Computing and the 2015 ACM International SymposiumonWearable Computers UbiComp and ISWC 2015 pp 349ndash352September 2015
[12] Y Nakamura T Kanehira Y Arakawa and K YasumotoldquoSenStick 2 Ultra tiny all-in-one sensor with wireless chargingrdquoinProceedings of the 2016ACM International Joint Conference onPervasive and Ubiquitous Computing UbiComp 2016 pp 337ndash340 September 2016
[13] ldquoSensortag cc2541 (texas instruments)rdquo httpwwwticomtoolcc2541dk-sensor
[14] ldquoIot smart modulerdquo httpwwwalpscomjiotsmart[15] ldquoTsnd121 (atr-promotions)rdquo httpwwwatr-pcomproducts
TSND121html[16] ldquoWax9 (axivity)rdquo httpaxivitycomproductwax9[17] Uk biobank ldquoUk biobankrdquo httpwwwukbiobankacuk[18] J C Seidell and J Halberstadt ldquoThe global burden of obesity
and the challenges of preventionrdquo Annals of Nutrition andMetabolism vol 66 supplement 2 pp 7ndash12 2015
[19] ldquoWeltThe smart belt for fashion and healthrdquo httpswwwwelt-corpcom
[20] ldquoBelty power The belt your really deserverdquo httpswwwbeltyco
[21] H Nam J-H Kim and J-I Kim ldquoSmart Belt A wearabledevice for managing abdominal obesityrdquo in Proceedings of theInternational Conference on Big Data and Smart ComputingBigComp 2016 pp 430ndash434 January 2016
[22] K Moriya M Fujimoto Y Arakawa and K Yasumoto ldquoIndoorlocalization based on distance-illuminance model and activecontrol of lighting devicesrdquo in Proceedings of the 2016 Interna-tional Conference on Indoor Positioning and Indoor NavigationIPIN 2016 pp 1ndash6 October 2016
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal of
Volume 201
Submit your manuscripts athttpswwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
Journal of Sensors 15
Typical signal of airso gun
Hold a gun Ready
Shoot
1 2 3 4 5 6 70
00
30
60
91
21
51
82
12
42
73
03
33
63
94
24
54
85
15
45
76
06
36
66
97
27
57
88
18
48
79
09
39
69
910
210
510
811
111
411
712
012
312
612
913
2
minus600
minus400
minus200
0
200
400
600
800
acc_xacc_yacc_z
gyro_xgyro_y
acc_synthesis gyro_zgyro_synthesis
Figure 16 Time series of sensing data in shooting an airsoft gun
Sensorized pencil
Figure 17 Sensorized pencil
Sensorized name tagVisualization of sensor data
Figure 18 Sensorized name tag
Sensorized spoon Digital fishing game
Figure 19 Sensorized spoon
Table 9 Questionnaire results about SenStick
Question contents ScoreQ1 Do you think that SenStick will be useful for IoTresearch 49
Q2 Do you think that SenStick will be useful for IoTeducation 47
Q3 Do you want to use SenStick again 49Q4 Do you think that SenStick was easy to use 41Q5 Are you satisfied with specification of SenStick 41
hardware software and 3D case data The main purpose ofour platform is to simplify the sensorization of our personalbelongings
The hardware of SenStick is maybe one of the worldrsquossmallest sensor boards which includes 8 kinds of sensorsand BLE In addition to the hardware we also developedsupport software for managing monitoring recording andviewing the hardware By providing 3D case data all theresearchers can measure every tiny activity in daily life easilyand efficiently For an educational use it is also useful forunderstanding sensors and prototyping the desired systemincluding tiny censored things
Conflicts of Interest
The authors declare that there are no conflicts of interestregarding the publication of this paper
Acknowledgments
This work is supported in part by JST PRESTO and JSPSKAKENHI (Grant no 17J10021) and ldquoResearch and Devel-opment of Innovative Network Technologies to Create theFuturerdquo the Commissioned Research of National Instituteof Information and Communications Technology (NICT)Japan
16 Journal of Sensors
References
[1] O Brdiczka M Langet J Maisonnasse and J L CrowleyldquoDetecting human behavior models from multimodal observa-tion in a smart homerdquo IEEE Transactions on Automation Scienceand Engineering vol 6 no 4 pp 588ndash597 2009
[2] J Chen A Kam J Zhang N Liu and L Shue ldquoBathroomactivity monitoring based on soundrdquo in Pervasive ComputingThird International Conference PERVASIVE 2005 MunichGermanyMay 8-13 2005 Proceedings vol 3468 ofLectureNotesin Computer Science pp 47ndash61 Springer Berlin Germany2002
[3] G Bauer K Stockinger and P Lukowicz ldquoRecognizing theUse-Mode of Kitchen Appliances from Their Current Consump-tionrdquo in Smart Sensing and Context vol 5741 of Lecture Notesin Computer Science pp 163ndash176 Springer Berlin HeidelbergBerlin Heidelberg 2009
[4] K Ouchi and M Doi ldquoIndoor-outdoor activity recognition bya smartphonerdquo in Proceedings of the the 2012 ACM Conferencep 537 Pittsburgh Pennsylvania September 2012
[5] A Anjum and M U Ilyas ldquoActivity recognition using smart-phone sensorsrdquo in Proceedings of the IEEE 10th ConsumerCommunications and Networking Conference (CCNC rsquo13) pp914ndash919 IEEE January 2013
[6] O D Lara and M A Labrador ldquoA survey on human activityrecognition using wearable sensorsrdquo IEEE CommunicationsSurveys amp Tutorials vol 15 no 3 pp 1192ndash1209 2013
[7] ldquoJins memerdquo httpsjins-memecom[8] Y Matsuda T Ishioka T Hasegawa et al ldquoWaistonBelt A belt
for monitoring your real abdominal circumference foreverrdquo inProceedings of the ACM International Joint Conference on Perva-sive and Ubiquitous Computing and the 2015 ACM InternationalSymposium on Wearable Computers UbiComp and ISWC 2015pp 329ndash332 September 2015
[9] M Kurahashi K Murao T Terada and M Tsukamoto ldquoPer-sonal identification system based on rotation of toilet paperrollsrdquo in Proceedings of the IEEE International Conference onPervasive Computing and CommunicationsWorkshops (PerComWorkshops rsquo17) pp 521ndash526 IEEE Kona Hawaii USA March2017
[10] ldquoHapiforkrdquo httpswwwhapicom[11] Y Arakawa ldquoSenStick Sensorize every thingsrdquo in Proceedings of
the ACM International Joint Conference on Pervasive and Ubiq-uitous Computing and the 2015 ACM International SymposiumonWearable Computers UbiComp and ISWC 2015 pp 349ndash352September 2015
[12] Y Nakamura T Kanehira Y Arakawa and K YasumotoldquoSenStick 2 Ultra tiny all-in-one sensor with wireless chargingrdquoinProceedings of the 2016ACM International Joint Conference onPervasive and Ubiquitous Computing UbiComp 2016 pp 337ndash340 September 2016
[13] ldquoSensortag cc2541 (texas instruments)rdquo httpwwwticomtoolcc2541dk-sensor
[14] ldquoIot smart modulerdquo httpwwwalpscomjiotsmart[15] ldquoTsnd121 (atr-promotions)rdquo httpwwwatr-pcomproducts
TSND121html[16] ldquoWax9 (axivity)rdquo httpaxivitycomproductwax9[17] Uk biobank ldquoUk biobankrdquo httpwwwukbiobankacuk[18] J C Seidell and J Halberstadt ldquoThe global burden of obesity
and the challenges of preventionrdquo Annals of Nutrition andMetabolism vol 66 supplement 2 pp 7ndash12 2015
[19] ldquoWeltThe smart belt for fashion and healthrdquo httpswwwwelt-corpcom
[20] ldquoBelty power The belt your really deserverdquo httpswwwbeltyco
[21] H Nam J-H Kim and J-I Kim ldquoSmart Belt A wearabledevice for managing abdominal obesityrdquo in Proceedings of theInternational Conference on Big Data and Smart ComputingBigComp 2016 pp 430ndash434 January 2016
[22] K Moriya M Fujimoto Y Arakawa and K Yasumoto ldquoIndoorlocalization based on distance-illuminance model and activecontrol of lighting devicesrdquo in Proceedings of the 2016 Interna-tional Conference on Indoor Positioning and Indoor NavigationIPIN 2016 pp 1ndash6 October 2016
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal of
Volume 201
Submit your manuscripts athttpswwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
16 Journal of Sensors
References
[1] O Brdiczka M Langet J Maisonnasse and J L CrowleyldquoDetecting human behavior models from multimodal observa-tion in a smart homerdquo IEEE Transactions on Automation Scienceand Engineering vol 6 no 4 pp 588ndash597 2009
[2] J Chen A Kam J Zhang N Liu and L Shue ldquoBathroomactivity monitoring based on soundrdquo in Pervasive ComputingThird International Conference PERVASIVE 2005 MunichGermanyMay 8-13 2005 Proceedings vol 3468 ofLectureNotesin Computer Science pp 47ndash61 Springer Berlin Germany2002
[3] G Bauer K Stockinger and P Lukowicz ldquoRecognizing theUse-Mode of Kitchen Appliances from Their Current Consump-tionrdquo in Smart Sensing and Context vol 5741 of Lecture Notesin Computer Science pp 163ndash176 Springer Berlin HeidelbergBerlin Heidelberg 2009
[4] K Ouchi and M Doi ldquoIndoor-outdoor activity recognition bya smartphonerdquo in Proceedings of the the 2012 ACM Conferencep 537 Pittsburgh Pennsylvania September 2012
[5] A Anjum and M U Ilyas ldquoActivity recognition using smart-phone sensorsrdquo in Proceedings of the IEEE 10th ConsumerCommunications and Networking Conference (CCNC rsquo13) pp914ndash919 IEEE January 2013
[6] O D Lara and M A Labrador ldquoA survey on human activityrecognition using wearable sensorsrdquo IEEE CommunicationsSurveys amp Tutorials vol 15 no 3 pp 1192ndash1209 2013
[7] ldquoJins memerdquo httpsjins-memecom[8] Y Matsuda T Ishioka T Hasegawa et al ldquoWaistonBelt A belt
for monitoring your real abdominal circumference foreverrdquo inProceedings of the ACM International Joint Conference on Perva-sive and Ubiquitous Computing and the 2015 ACM InternationalSymposium on Wearable Computers UbiComp and ISWC 2015pp 329ndash332 September 2015
[9] M Kurahashi K Murao T Terada and M Tsukamoto ldquoPer-sonal identification system based on rotation of toilet paperrollsrdquo in Proceedings of the IEEE International Conference onPervasive Computing and CommunicationsWorkshops (PerComWorkshops rsquo17) pp 521ndash526 IEEE Kona Hawaii USA March2017
[10] ldquoHapiforkrdquo httpswwwhapicom[11] Y Arakawa ldquoSenStick Sensorize every thingsrdquo in Proceedings of
the ACM International Joint Conference on Pervasive and Ubiq-uitous Computing and the 2015 ACM International SymposiumonWearable Computers UbiComp and ISWC 2015 pp 349ndash352September 2015
[12] Y Nakamura T Kanehira Y Arakawa and K YasumotoldquoSenStick 2 Ultra tiny all-in-one sensor with wireless chargingrdquoinProceedings of the 2016ACM International Joint Conference onPervasive and Ubiquitous Computing UbiComp 2016 pp 337ndash340 September 2016
[13] ldquoSensortag cc2541 (texas instruments)rdquo httpwwwticomtoolcc2541dk-sensor
[14] ldquoIot smart modulerdquo httpwwwalpscomjiotsmart[15] ldquoTsnd121 (atr-promotions)rdquo httpwwwatr-pcomproducts
TSND121html[16] ldquoWax9 (axivity)rdquo httpaxivitycomproductwax9[17] Uk biobank ldquoUk biobankrdquo httpwwwukbiobankacuk[18] J C Seidell and J Halberstadt ldquoThe global burden of obesity
and the challenges of preventionrdquo Annals of Nutrition andMetabolism vol 66 supplement 2 pp 7ndash12 2015
[19] ldquoWeltThe smart belt for fashion and healthrdquo httpswwwwelt-corpcom
[20] ldquoBelty power The belt your really deserverdquo httpswwwbeltyco
[21] H Nam J-H Kim and J-I Kim ldquoSmart Belt A wearabledevice for managing abdominal obesityrdquo in Proceedings of theInternational Conference on Big Data and Smart ComputingBigComp 2016 pp 430ndash434 January 2016
[22] K Moriya M Fujimoto Y Arakawa and K Yasumoto ldquoIndoorlocalization based on distance-illuminance model and activecontrol of lighting devicesrdquo in Proceedings of the 2016 Interna-tional Conference on Indoor Positioning and Indoor NavigationIPIN 2016 pp 1ndash6 October 2016
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal of
Volume 201
Submit your manuscripts athttpswwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal of
Volume 201
Submit your manuscripts athttpswwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of