SSSVVVMMM을을을이이이용용용한한한다다다중중중생생생체체체인인인식식식 …repository.kmou.ac.kr/bitstream/2014.oak/8477/1/000002174450.pdf · 제5장

SSSVVVMMM

AAA SSStttuuudddyyyooonnnDDDeeesssiiigggnnnaaannndddIIImmmpppllleeemmmeeennntttaaatttiiiooonnnooofffMMMuuullltttiii---MMMooodddaaalllBBBiiiooommmeeetttrrriiicccsssRRReeecccooogggnnniiitttiiiooonnnSSSyyysssttteeemmm uuusssiiinnngggSSSVVVMMM

222000000777 888

222000000777 888

-i-

................................................................................................................i ......................................................................................................iii ...........................................................................................................vAbstract........................................................................................................vi 1 .............................................................................................1 2 .....................................................................................5

2.1 ...............................................52.2 ...............................................62.3 ...............................122.4 RFID ...............................18

3 SVM ..................................................................................233.1 ..........................................................................................233.2 ......................................................................253.3 SVM ....................................................................303.4 SVM ................................................................35

4 ...........434.1 ..........................................................434.2 ..................................444.3 ..................................464.4 RFID ..........................................48

5 .........515.1 .....................................................51

-ii-

5.2 ...........................................535.3 ...........................................615.4 RFID ..................................................67

6 .....................................................................696.1 RFID ......................................................696.2 ......................................................................706.3 ......................................................................726.4 ..............................................736.5 SVM ............................................836.6 SVM ...........87

7 ...........................................................................................91 .................................................................................................93 ..........................................................................................................100

-iii-

2.1 ...................................5 2.2 / .........................8 2.3 // ...8 2.4 / .........................9 2.5 .................................................10 2.6 ........................................................................11 2.7 .................................13 2.8 .................................15 2.9 ........17 2.10 RFID .............................................................19 3.1 ..................................................................29 3.2 .......................................31 3.3 .............................33 4.1 ..................................................44 4.2 ......................................................45 4.3 .......................................45 4.4 ..........................................................46 4.5 .......................................47 4.6 RFID ....................................................50 4.7 RFID .......................50 5.1 SVM

...............................................................................52

-iv-

5.2 ...................................54 5.3 ...................................61 5.4 ..................................................63 6.1 ...........................71 6.2 ..........................................71 6.3 ...........................72 6.4 ................................................74 6.5 croppedpoint ........................76 6.6 sectorizedpoint ....................77 6.7 normalizedpoint ..................78 6.8 gaborfilter ..................78 6.9 convolution ............................79 6.10 .................................80 6.11 ........................................................................80 6.12 SVM .....................................................83 6.13 Linear ..............................84 6.14 Polynomial .....................84 6.15 GaussianRBF ...............85 6.16 Linearspline .................85 6.17 Bspline ............................86 6.18 ExponentialRBF .........86 6.19 SVM ..................................................89

-v-

4.1 ...............................................................................48 4.2 RFID(S6700) ..............................49 6.1 RFID ....................................................................70 6.2 FRR ..................................................................74 6.3 FAR .................................................................75 6.4 FRR ..................................................................81 6.5 FAR .................................................................82 6.6 ................................................................88 6.7 .............89

-vi-

AAAbbbssstttrrraaacccttt

Thisstudyintendstodesignandimplementthemulti-modalbiometric system by applying the RFID recognition systempresentlyhasbeengreatlyissuedintheubiquitoussocietytothemulti-modalbiometric system using the fingerprint and thespeechwidelyusedinthebiometrics,inordertoimproveFARandFRRoftheexistingsinglebiometricsystem.The recognition system is designed with TIs DSP TMS-

320C32,TMS320VC5509andAVRsothatitcanbeimplementedasembeddedsubsystemsandisconnectedtotheRFID systeminserialwhilethefingerprintrecognitionsystem andthespeechrecognitionsystem areconnectedinparallel.The hardware is built with the algorithms excellent in

real-time forefficientreal-time implementation.Among them,SVM algorithm,which is the mostessentialin this study,distinguishes the registered from the impostors by using thedistancevalueobtainedtworecognitionsystems.Thisintegratesystem improvesFARandFRR.Inthisstudy,weperformedtheSVM performancetestthenumberofvectorswithinthesupportvectoristheleast,whichis27(10.3%),withtheBsplinekernelfunctionamong7kernelfunctions;therefore,itisverifiedthatthekernelfunctionofthisstudygivesthebestefficiency.Wecompared with other algorithms(weight sum,fuzzy integral,

-vii-

decisiontree)inordertoevaluatetheperformanceofthemulti-modalbiometricsystem usingthelinearBslpinefunction,andtheresultshowsthatitismuchmoreefficientwithmaximum 12%difference.

-1-

111

, , . (token-based) (knowledge-based) . ,

,ID , ,RFID(RadioFrequencyIdentification),, , , RFID . ,PIN(PersonalIdentificationNumber). 2 , , ,, . , (biometricidentification) .

. , , 2

-2-

. (universality) ,

(uniqueness), (permanence), (collectability) . (performance), (acceptability), (circumvention) . ,,,,

, , 50% , ,,, . , . 1990

PC ,2000 . (DSP[1][2],ARM ) 90% . , . , ,

-3-

, , .

[3]-[14] . . , ,

, SVM(SupportVectorMachines) , SVM .

, ,,RFID . 2 RFID TI(32Bit)AVR(8bit) , SVM [15]-[29] . .1 ,2

RFID[30]-[39] ,3 SVM .4 5

-4-

, 4 ,5 . 6 , Matlab[40][41] (simulation) , . 7 .

-5-

222

222...111

, .

, , , ,

, , , ,

, , , ,

((((DNA)DNA)DNA)DNA)

, , , , ,,,,

, , , ,

, , , ,

, , , ,

((((DNA)DNA)DNA)DNA)

, , , , ,,,,

2.1 Fig.2.1Componentofhumanforbiometricrecognition

-6-

,,,,,,,, ,DNA , . ,, ,,,,, . 2.1 , .

, , , 5 .

222...222

. (multiple sensors), (multiplebiometrics), (multipleunitsofthesamebiometric),(multipleinstancesofthesamebiometric), (multiplematchers) , 5 .

-7-

222...222...111

. 3 ,,, . 3 3 .

222...222...222

. ,/,//,/,/,// . .

(1) +

2.2 ACTS-M2VTSEuropeanproject , () .

-8-

////

////

2.2/ Fig.2.2Exampleofmulti-modalbiometricrecognitionsystem

usingface/speech

(2) + +

2.3// Fig.2.3Exampleofmulti-modalbiometricrecognitionsystem

usingface/speech/lip-reading

-9-

2.3 // , / (lip-reading) ,, .

(3) +

2.4 .

EigenfaceEigenfaceEigenfaceEigenface

MinutiaeMinutiaeMinutiaeMinutiae









2.4/ Fig.2.4Exampleofmulti-modalbiometricrecognitionsystem

usingface/fingerprint

-10-

, .

222...222...333

, . , , 10 .

222...222...444

, , .

2.5 Fig.2.5Exampleofmultipleinstancesofthesamebiometric

offacerecognition

-11-

2.5 , .

222...222...555

, 2.6 .

Minutiae Minutiae Minutiae Minutiae 1. Orientation 1. Orientation 1. Orientation 1. Orientation 2. 2. 2. 2. 3. 3. 3. 3. 4. 4. 4. 4. MiuntiaeMiuntiaeMiuntiaeMiuntiae

Texture Texture Texture Texture 1. 1. 1. 1. 2. 2. 2. 2. 3. 3. 3. 3. 4. AAD4. AAD4. AAD4. AAD

1. 1. 1. 1. FingerCodeFingerCodeFingerCodeFingerCode 2. 2. 2. 2.

1. 1. 1. 1. 2. ID 2. ID 2. ID 2. ID 3. 3. 3. 3.

1. 1. 1. 1. 2. 2D 2. 2D 2. 2D 2. 2D


Texture Texture Texture Texture 1. 1. 1. 1. 2. 2. 2. 2. 3. 3. 3. 3. 4. AAD 4. AAD 4. AAD 4. AAD


Texture Texture Texture Texture 1. 1. 1. 1. 2. 2. 2. 2. 3. 3. 3. 3. 4. AAD4. AAD4. AAD4. AAD

1. 1. 1. 1. FingerCodeFingerCodeFingerCodeFingerCode 2. 2. 2. 2.

1. 1. 1. 1. 2. ID 2. ID 2. ID 2. ID 3. 3. 3. 3.

1. 1. 1. 1. 2. 2D 2. 2D 2. 2D 2. 2D


Texture Texture Texture Texture 1. 1. 1. 1. 2. 2. 2. 2. 3. 3. 3. 3. 4. AAD 4. AAD 4. AAD 4. AAD

2.6 Fig.2.6Exampleofmultiplematchers

-12-

2 , 2 .

222...333

(weightsum rule)[42] (fuzzyintegral)[43] , (decisiontree)[44] . SVM , 3 .

222...333...111

2 2 , . 2.7 , .

-13-

2.7 Fig.2.7Multi-modalbiometricrecognitionusing

weightsum rule

(2.1) , , , .

(2.1)

(2.2) (threshold)t / . 0 , .

-14-

(2.2)

222...333...222

, . (sugeno) (choquet) . min-max , . 2.8

, (2.3) , g(Ai) ., , , . ()

() .

(2.3)

-15-

2.8 Fig.2.8Multi-modalbiometricrecognitionusing

fuzzyintegral

222...333...333

. . ,, , . , .

-16-

. , .

. .

if-then-else., 2.9 .

. 2.9 9 , if-then-else .

if( max)then =

elseif(min

-17-

2.9 Fig.2.9Classificationoftheimpostor/registeredusing

decisiontree

elseif( >max, >max)then =

elseif(

-18-

elseif( >max,

-19-

((((+EPC)+EPC)+EPC)+EPC)

(LAN)

(LAN)

ONS + PML ONS + PML ONS + PML ONS + PML

SAVANTSAVANTSAVANTSAVANT

((((+EPC)+EPC)+EPC)+EPC)

(LAN)

(LAN)

ONS + PML ONS + PML ONS + PML ONS + PML

SAVANTSAVANTSAVANTSAVANT

2.10RFID Fig.2.10 Componentdiagram ofRFIDsystem

, IC . IC () .

. ASK,FSK,PSK , . ( UHF) SS(SpreadSpectrum) CDMA

-20-

. SS ,ASK .

,CRC , Aloha CSMA , . .RFID EPC(Electronic

ProductCode) RFID RFID Savant, PML(PhysicalMark-upLanguage),PML ONS(ObjectNameService) .

222...444...222RRRFFFIIIDDD

,, . . RFID .

-21-

(1)

. , , . . . . . . , .

(2)

, . , , .

-22-

. . . , . .

-23-

333 SSSVVVMMM

. (fusion) (classification) . ,

(weightsum rule), (fuzzyintegral), (decisiontree) . [45]-[47] SVM , SVM , SVM .

333...111

. . (quantity) (quality) . (finite) ,(sample)

-24-

, .

(overfitting) . . SVM AT&T Bell

Vapnik (classification) (regression) , . SRM(StructuralRiskMinimisation) , ERM(EmpiricalRiskMinimisation) .SRM ERM (risk) (bound) , .SVM , . SVM .

, , . ERM ,

-25-

. , , .

333...222 (((ssstttaaatttiiissstttiiicccaaalllllleeeaaarrrnnniiinnngggttthhheeeooorrryyy)))

. {+1,-1} , (3.1) ., +1 -1 .

(3.1)

, , . (3.2) .

(3.2)

-26-

, . .

333...222...111

.ERM l . (3.3) , .

(3.3)

, .

lim (3.4)

lim (3.5)

-27-

.VC (3.6) , VC .

lim (3.6)

333...222...222VVVCCC

. , VC . VC f , (3.7) , (confidence), .

(3.7)

-28-

VC (3.7) VC .VC R[f] Remf[f] VC .

333...222...333

ERM . . Vapnik Cherbonenkis 1974 VC SRM ., .SRM

, . S

-29-

, ,

,VC

. (3.7) . 3.1 , .

(empricial risk)

(bound on the risk)

(confidence)

h1 h* hn h

S1 S* Sn

(empricial risk)

(bound on the risk)

(confidence)

h1 h* hn h

S1 S* Sn

3.1 Fig.3.1 Boundontherisk

,VC

-30-

. VC , VC , 3.1 h*. 3.3 3.4 SVM [48]

.

333...333 SSSVVVMMM

SVM (xi i ,di ) .=+1 ()=-1 . (3.8) .

(3.8)

, ,bias. SVM

(3.9) .

fordi=+1 (3.9) fordi=-1

, (3.8) (hyperplane)

-31-

() .SVM

, . 3.2 2 .

3.2 Fig.3.2Geometricalstructureofoptimum hyperplane

. (weight) bias , (3.10) .

(3.10)

(3.10) (discriminantfunction) , (3.11) .

-32-

(3.11)

, (3.12) .

(3.12)

, . (positive) , (negative) . g(xp)=0 (3.13) .

(3.13)

, bo/||wo|| , bo>0 ,bo

-33-

3.3 . , wo bo . 3.3 (wo,bo) (3.14) . (3.9) , (3.14)

wo bo , (canonical) .

fordi=+1 (3.14)

fordi=-1

3.3 Fig.3.3Findingparameterforoptimum hyperplane

(3.14) {xi,di}(supportvector) ,

-34-

(machine) . , .

,d(s)=+1 x(s) , (3.15) .

(3.15)

(3.13) x(s) (3.16) .

(3.16)

(+) x(s) ,(-) x(s) . , (3.16) (3.17) .

(3.17)

(3.17) w Euclideannorm .

-35-

333...444 SSSVVVMMM

SVM , (nonlineardecisionsurface) , . SVM x , . ., . SVM .

333...444...111

x mo ,

. m1 . , (3.18) .

-36-

(3.18)

,b bias, (3.19) .

(3.19)

w 0 biasb x . (3.19) , wj , (3.20) .

(3.20)

(3.21) .

(3.21)

x , (3.22) .

-37-

(3.22)

(3.22) [ A] (A.4) , (3.23) .

(3.23)

i (3.18) (3.19) (3.24) .

(3.24)

(3.24) x i . ,Mercer[ B] (3.25) .

(3.25)

-38-

(3.25) , (3.26) .

(3.26)

, . (3.25) (3.24) , (3.27) .

(3.27)

333...444...222SSSVVVMMM

(3.25) . , SVM . (1) bias Q()

. .

-39-

(3.28) N*N K ij .

(3.28)

, o,i (3.23) ., w , (3.29) . , biasbo .

(3.29)

, (Lagrange) .

.

(1)

, (2)

-40-

333...444...333

(polynomial) ,RBF(RadialBasisFunction) , (multi-layerperceptron) 9 . (Gaussian)RBF ,(exponential)RBF ,B-splines ,(multidimensional) [29]. 4 ,(Linear)Splines [29], 6 .

(1)

(dotproduct) , .

(3.30)

(2) RBF

RBF , RBF

-41-

. .

(3.31)

(3) RBF

RBF .

(3.32)

(4) Splines

Splines .k Splines(s N ) (3.33) .

(3.33)

Splines([0,1]) (3.34) .

(3.34)

k=1 , (3.35) .

-42-

(3.35)

(5)Bsplines

Splines , ( [-1,1]) (3.36) .

(3.36)

(6)

,(tensorproduct) (3.37) .

(3.37)

-43-

444

. , ,RFID , ,5 .

444...111

4.1 SVM PC , PC RS-232(serial), (parallel) , RFID RS-232 . , (SVM) , (DeadBolt) .

-44-

4.1 Fig.4.1Overallcomponentdiagram ofcontrolboard

444...222

4.2[ C] CPU TI TMS320VC5509 , DSP SDRAM . 256 1MB , 2MB SDRAM . AuthenTec AFS-8500 . DSP 9696 . 4.3 .

-45-

MT28F800B3WGMT28F800B3WGMT28F800B3WGMT28F800B3WG----9 9 9 9 (Flash (Flash (Flash (Flash ))))

8Mb*16 = 512kWord8Mb*16 = 512kWord8Mb*16 = 512kWord8Mb*16 = 512kWord

TMS320VC5509TMS320VC5509TMS320VC5509TMS320VC5509

CE1CE1CE1CE1

CE2CE2CE2CE2

IS42S16100IS42S16100IS42S16100IS42S16100----7T 7T 7T 7T (SDRAM (SDRAM (SDRAM (SDRAM ))))

16Mb*16 = 1MWord16Mb*16 = 1MWord16Mb*16 = 1MWord16Mb*16 = 1MWord

CE0CE0CE0CE0

AFSAFSAFSAFS----8500850085008500FingerprintFingerprintFingerprintFingerprint

SensorSensorSensorSensor

SPISPISPISPI

MT28F800B3WGMT28F800B3WGMT28F800B3WGMT28F800B3WG----9 9 9 9 (Flash (Flash (Flash (Flash ))))

8Mb*16 = 512kWord8Mb*16 = 512kWord8Mb*16 = 512kWord8Mb*16 = 512kWord

TMS320VC5509TMS320VC5509TMS320VC5509TMS320VC5509

CE1CE1CE1CE1

CE2CE2CE2CE2

IS42S16100IS42S16100IS42S16100IS42S16100----7T 7T 7T 7T (SDRAM (SDRAM (SDRAM (SDRAM ))))

16Mb*16 = 1MWord16Mb*16 = 1MWord16Mb*16 = 1MWord16Mb*16 = 1MWord

CE0CE0CE0CE0

AFSAFSAFSAFS----8500850085008500FingerprintFingerprintFingerprintFingerprint

SensorSensorSensorSensor

SPISPISPISPI

4.2 Fig.4.2Componentdiagram offingerprintrecognitionboard

4.3 Fig.4.3Implementalcontrolandfingerprintrecognitionboard

-46-

444...333

. [ D] TI DSPTMS320C32(30MIPS) SRAM 61C256*4(128Kbyte) , 64KbyteROM 27C512 .

STRB0STRB0STRB0STRB0


IOSTRBIOSTRBIOSTRBIOSTRB

ROMROMROMROM(27c512)(27c512)(27c512)(27c512)

TMS320C32TMS320C32TMS320C32TMS320C32

RAM(61256)RAM(61256)RAM(61256)RAM(61256)

RAM(61256)RAM(61256)RAM(61256)RAM(61256)

RAM(61256)RAM(61256)RAM(61256)RAM(61256)

RAM(61256)RAM(61256)RAM(61256)RAM(61256)

RAM(61256)RAM(61256)RAM(61256)RAM(61256)

RAM(61256)RAM(61256)RAM(61256)RAM(61256)

RAM(61256)RAM(61256)RAM(61256)RAM(61256)

RAM(61256)RAM(61256)RAM(61256)RAM(61256)

GAL22V10GAL22V10GAL22V10GAL22V10 PPI8255PPI8255PPI8255PPI8255

CODECCODECCODECCODEC(CS4218)(CS4218)(CS4218)(CS4218)

LCDLCDLCDLCD

SerialSerialSerialSerial



IOSTRBIOSTRBIOSTRBIOSTRB

ROMROMROMROM(27c512)(27c512)(27c512)(27c512)

TMS320C32TMS320C32TMS320C32TMS320C32

RAM(61256)RAM(61256)RAM(61256)RAM(61256)

RAM(61256)RAM(61256)RAM(61256)RAM(61256)

RAM(61256)RAM(61256)RAM(61256)RAM(61256)

RAM(61256)RAM(61256)RAM(61256)RAM(61256)

RAM(61256)RAM(61256)RAM(61256)RAM(61256)

RAM(61256)RAM(61256)RAM(61256)RAM(61256)

RAM(61256)RAM(61256)RAM(61256)RAM(61256)

RAM(61256)RAM(61256)RAM(61256)RAM(61256)

RAM(61256)RAM(61256)RAM(61256)RAM(61256)

RAM(61256)RAM(61256)RAM(61256)RAM(61256)

RAM(61256)RAM(61256)RAM(61256)RAM(61256)

RAM(61256)RAM(61256)RAM(61256)RAM(61256)

RAM(61256)RAM(61256)RAM(61256)RAM(61256)

RAM(61256)RAM(61256)RAM(61256)RAM(61256)

RAM(61256)RAM(61256)RAM(61256)RAM(61256)

RAM(61256)RAM(61256)RAM(61256)RAM(61256)

GAL22V10GAL22V10GAL22V10GAL22V10 PPI8255PPI8255PPI8255PPI8255

CODECCODECCODECCODEC(CS4218)(CS4218)(CS4218)(CS4218)

LCDLCDLCDLCD

SerialSerialSerialSerial

4.4 Fig.4.4Componentdiagram ofspeechrecognitionboard

MPSD-PP (emulator) JTAG , 16bit (CS4218)

-47-

/ ,XF0 XF1 I/O CPU . DSP INT2 INT3 . 4.4 , .

TMS320C32 DSP . DSP MFCC DTW . LCD , . 4.5 .

4.5 Fig.4.5Implementalcontrolandspeechrecognitionboard

-48-

444...444RRRFFFIIIDDD

RFID ,RF, , . 4.1 .

4.1 Table4.1Specificationofreader

RF 13.56MHz,FCC CECompliant

4.5 5.0DC,5VDC (6 9V ) 70mA @5V 178*88*11mmRF 26kbpsISO15693-3,106kbpsISO14443-A UART,RS-232

9600-115200baudrate 50 & RF 100mW @ 5VRead 100mm FlashMCU 256byte 8k

LED2(red/green), 15

-49-

4.1 13.56MHz RFID , ISO15693-3 ISO14443-A,(AVR PC) UART RS-232, 96001152000baudrate. 4.6 ,

ATMEL8L ,RFID ATMEL16L . RFID S6700(RI-R6C-001A) 4.2 . 4.7 RFID

, Atmel16L LCD, [ E]. ISO 14443-A ISO 15693 , .

4.2 RFID(S6700) Table4.2MajorfunctionandperformanceofRFID(S6700)

Tag ISO15693-2

ISO14443-2 50 200mW RF

TransparentMode

-50-

IO IO IO IO

ATMEL8LATMEL8LATMEL8LATMEL8L

RSRSRSRS----232C232C232C232C5.1~1

10056.13

30~10Q

%1050

==

=

VSWR

kHzF

ohm

res

S6700S6700S6700S6700(RI(RI(RI(RI----R6CR6CR6CR6C----001A)001A)001A)001A)

(ATMEL16L)(ATMEL16L)(ATMEL16L)(ATMEL16L)

RSRSRSRS----232C232C232C232C

IO IO IO IO

ATMEL8LATMEL8LATMEL8LATMEL8L

RSRSRSRS----232C232C232C232C5.1~1

10056.13

30~10Q

%1050

==

=

VSWR

kHzF

ohm

res

S6700S6700S6700S6700(RI(RI(RI(RI----R6CR6CR6CR6C----001A)001A)001A)001A)

(ATMEL16L)(ATMEL16L)(ATMEL16L)(ATMEL16L)

RSRSRSRS----232C232C232C232C

4.6RFID Fig.4.6Componentdiagram ofRFIDrecognitionboard

4.7 RFID Fig.4.7OverallfigureofimplementalRFIDrecognitionsystem

-51-

555

. 4 , RFID .

555...111

3 . RFID . [0,1] (opinionvector) . FAR FRR . SVM , SVM .

-52-

SVM 5.1 .

5.1SVM

Fig.5.1Softwareblockdiagram ofproposedmulti-modalbiometricsystem usingSVM

3 ,RFID

-53-

, . RFID (SVM) , . RFID ,4 . (matching), , (decisionspace) , 3 SVM .

555...222

[49]-[56] (fingercode) . 5.2 . 2

. , .

, 16, 5 . (wedge-ring) . , (normalization) , 0 180

-54-

22.5 8 (gabor) (convolution). . , . .

1.1.1.1.2.2.2.2.

1111 2222

1.1.1.1.2.2.2.2.

1111 2222

5.2 Fig.5.2Overallcomponentdiagram offingerprint

recognitionalgorithm

-55-

555...222...111

, (gradient) , .

(1)WW(8,8) .

(2) (sobel) (gradient), .

(3) (orientation) .

(5.1)

(5.2)

(5.3)

-56-

555...222...222

(poincare) (max-curvature),(sine-component) 3 . , .

(1)5.2.1 .

(2) LPF . . . W 2 low passfilter.

(5.4)

(5.5)

(5.6)

(5.7)

-57-

(5.8)

(3) .

(5.9)

(4) A .

(5) A(i,j) .A(i,j) .

(5.10)

(i,j)= (i,j)

RRRR IIIIIIII

RRRR IIII RRRR IIII

RRRR IIIIIIII

RRRR IIII RRRR IIII

(6) w',w''(w''

-58-

555...222...333 (((gggaaabbbooorrrfffiiilllttteeerrr)))

(ridges) (valleys) (local) , , . , , . , (half) . 4 . .

(1) 16 (Sectorize) .

(2) (Normalization) .

(5.11)

(3) .

-59-

555...222...444

, , . . , . , . 8 , , (5.13) .

(5.13)

(5.12)

,f ,x x,y y .

-60-

, ,(x,y) , i .

555...222...555 ::: (((eeeuuucccllliiidddeeeaaannndddiiissstttaaannnccceee)))

, . , . 0,22.5,45,67.5,90 5 . (5.14) , .

(5.14)

(5.15) , . ,

.

-61-

(5.15)

555...333

5.3.

PCM PCM PCM PCM

0000 1111

????

6666????

20202020????

5555

YesYesYesYes

YesYesYesYes

YesYesYesYes

NoNoNoNo

NoNoNoNo

1111

) ) ) ) , , , , ,,,,

DTWDTWDTWDTW DBDBDBDB

< 30 ?< 30 ?< 30 ?< 30 ?

YesYesYesYes

NoNoNoNo

(FFT)(FFT)(FFT)(FFT)

----

(DCT)(DCT)(DCT)(DCT)

MFCC 12MFCC 12MFCC 12MFCC 12

PCM PCM PCM PCM

0000 1111

????

6666????

20202020????

5555

YesYesYesYes

YesYesYesYes

YesYesYesYes

NoNoNoNo

NoNoNoNo

1111

) ) ) ) , , , , ,,,,

DTWDTWDTWDTW DBDBDBDB

< 30 ?< 30 ?< 30 ?< 30 ?

YesYesYesYes

NoNoNoNo

(FFT)(FFT)(FFT)(FFT)

----

(DCT)(DCT)(DCT)(DCT)

MFCC 12MFCC 12MFCC 12MFCC 12

5.3 Fig.5.3Overallcomponentdiagram ofspeech

recognitionalgorithm

-62-

HMM(Hidden Markov Models),DTW(DynamicTimeWrapping),VQ(VectorQuantization),GMM(GaussianMixtureModel) , GMM . GMM . [57]-[61] DTW .

555...333...111

. . , (ZCR:ZeroCrossingRate) (shorttimeenergy) .

, . , (5.16) .

(5.16)

-63-

(5.16) x(n) ,N ,E . . 100 ,

30 . 6 , 20 . 20 15 , 5 .

555...333...222

. . MFCC(Mel-FrequencyCepstrum Coefficient) .MFCC 5.2 .(pre-emphasis) ,(hammingwindow) ,FFT(FastFourierTransform) . 5.6 MFCC ,,DCT (DistanceCosineTransform) 12

-64-

, .

5.3.2.1

DC , (5.17) . S(n) , a 0.95 .

(5.17)

5.3.2.2

.x(n) , w(n) (5.19) . N . (5.18) .

(5.18)

(5.19)

-65-

5.2.2.3MFCC

.(Mel) , .Stevens Volkman 1000Hz 1000mel, 2000mel . . 1kHz 1kHz (logscale) . (5.20) . Fmel ,FHz .

(5.20)

FFT , 5.4 . , (5.21) . , . 12 .

-66-

1KHz1KHz1KHz1KHz2KHz2KHz2KHz2KHz 3KHz3KHz3KHz3KHz

1111

4KHz4KHz4KHz4KHz

5.4 Fig.5.4Trianglebandwidthfilterbankofmel-cepstrum

(5.21)

5.2.3

DTW . , . , M T =T(1),T(2),...,T(M)

N R=R(1),R(2),...,R(N) , d (5.22) .

(5.22)

R n T W(n)

-67-

(localdistance),DTW (m,n) m=W(n) . ,

ITAKURA , . . DTW . DTW , (distancevalue) .


(Proximity) ISO14443-A ISO 15693[62-67] .

5.3.1ISO14443-A

ISO 14443 . 4 Part

-68-

,Part1 ,Part2 ,Part3 . Part4 .

5.3.2ISO15693

ISO 15693 . 1m . 4 Part ,Part1 ,Part2 ,Part3 . Part4 .

-69-

666

5 . RFID RFID , , . , , SVM Matlab .


RFID (reader) (tag) () RFID , RFID 4.6 RFID . ( )10 10 , 100 , 6.1 100%, . ,

-70-

20cm . 0~5Cm .

6.1RFID Table6.1TestofRFIDrecognitionrate

0cm 100% 0.1ms

5cm 100% 0.15ms

10cm 95% 0.52ms

15cm 67% 1.14ms

20cm 15%

666...222

6.1 . 10 3 30

. ( ) ,30 . (MFCC) 6.2 .

-71-

6.1 Fig.6.1Usingspeechdatabaseinspeechrecognition

6.2 Fig.6.2Databaseoftheextractedspeechfeaturepattern

-72-

666...333

.AuthenTec AFS-8500 3 , 10 . 6.3 , 10 30 , 30 . .

6.3 Fig.6.3Usingfingerprintdatabaseinfingerprintrecognition

-73-

666...444

2 . (FAR :FalseAcceptanceRatio). 1% 100 1 . . FAR . , (FRR:FalseRejectionRatio). . . FRR FAR .

666...444...111

6.4 SITK-32C DSP . 6.2 .

FRR 10 150 ,FRR 8.6% . 30 ,30 ,30

-74-

. 6.3 FAR , ,FAR 2.33% .

6.4 Fig.6.4Processofrealtestinspeechrecognition

6.2 FRR (/)Table6.2FRRtestofspeechrecognition

KJH 138 12 25.9RHS 145 5 23.5

-75-

6.3 FAR(/)Table6.3FARtestofspeechrecognition

KJH 0 150 120.4RHS 4 146 97.5JWI 0 150 130.8LSB 9 141 70.1KDH 3 147 92.3KJW 0 150 100.3MHG 0 150 111.7KJM 10 140 99.3SCW 8 142 120.2LCK 1 149 140.2

JWI 125 25 35.1LSB 140 10 24.7KDH 139 11 25.2KJW 135 15 30.2MHG 145 5 22.8KJM 119 31 36.1SCW 146 4 20.8LCK 139 11 24.4

-76-

666...444...222

, , Matlab , 5 Matlab . 6.5

croppedpoint .

6.5 croppedpointFig.6.5Croppedpointstepinprocessoffingerprint

preprocessing

-77-

6.6 .

6.6 sectorizedpointFig.6.6Sectorizedpointstepinprocessoffingerprint

preprocessing

6.7 , 6.8 .

-78-

6.7 normalizedpointFig.6.7Normalizedpointstepinprocessoffingerprint

preprocessing

6.8 gaborfilterFig.6.8Gaborfilterstepinprocessoffingerprintpreprocessing

-79-

22.5 , 6.9 convolution . 6.10 , . ,

, , , 6.11 2 .

6.9 convolutionFig.6.9Convolutionstepinprocessoffingerprintpreprocessing

-80-

6.10 Fig.6.10Featuringfingercodestepinprocessoffingerprint

preprocessing

6.11 Fig.6.11Resultoffingerprintrecognition

-81-

10 150 , FRR 6.4 . 10

, 10 ,10 .

6.4 FRR(/)Table6.4FRRtestoffingerprintrecognition


-82-

, 10 , 3.5% . FAR

6.5 , 1.7% ,SF FAR .

6.5 FARTable6.5FARtestoffingerprintrecognition


-83-

666...555 SSSVVVMMM

SVM SVM (kernelfunction) . 0 1 . 10 ,

5 , 15 20 300 . 6.12 () ,() , 7 .

6.12SVM Fig.6.12TrainingdatainSVM

-84-

Linear - :84(32.1%)

6.13Linear Fig.6.13Classifieddatainlinearkernelfunction

Polynomial - :85(32.4%)

6.14Polynomial Fig.6.14Classifieddatainpolynomialkernelfunction

-85-

GaussianRBF : :249 (95%)

6.15GaussianRBF Fig.6.15ClassifieddataingaussianRBFkernelfunction

Linearspline : :55 (21%)

6.16Linearspline Fig.6.16Classifieddatainlinearsplinekernelfunction

-86-

Bspline : :27(10.3%)

6.17Bspline Fig.6.17ClassifieddatainBsplinekernelfunction

ExponentialRBF : :52(19.8%)

6.18ExponentialRBF Fig.6.18ClassifieddatainexponentialRBFkernelfunction

-87-

(multidimensional) :

7 Bspline 27(10.3%) . , . (decisionboundary) , , . , SVM .

666...666 SSSVVVMMM

FAR FRR , . FAR,FRR . SVM 2

-88-

, , FAR FRR Bspline SVM FAR FRR . 6.2 6.5

FAR FRR FAR FRR, , , ,SVM 1500(10*150) . , , 6.6 .

6.6 Table6.6Settingvalueofeachalgorithm

- :16- :70% :30%

- :,

-min:10max:30

-min:30max:90

SVM 6.5

-89-

Bspline SVM , (parameter) 6.19 (SVM) .

6.19 SVM Fig.6.19SVM Structureinthispaper

6.7 FAR+FRR , FRR+FAR 8.4% [4] . 0.8%

-90-

, min,max . SVM ( , , ) .

6.7 Table6.7Comparativetestbetweensingleandmulti-modal

biometricrecognition

FFFAAARRR FFFRRRRRR FFFAAARRR+++FFFRRRRRR

( ) 2.33% 8.6% 10.93%

( ) 1.7% 3.5% 5.2%

( ) 5.33% 6.8% 12.13%

( ) 4.8% 3.6% 8.4%

( ) 0.8% 2.4% 2.4%

(SVM) 0% 0.13% 0.13%

-91-

777

(FAR/FRR) RFID . TI DSP

TMS320C32 TMS320VC5509 AVR , RFID , . . SVM 2 FAR FRR . SVM ,7 Bspline 27(10.3%) . Bspline ( , , ) 12% . SVM Bspline

-92-

FAR FRR .

, . .

-93-

[1]www.ti.com,"TMS320C32 GeneralPurpose User's guide",TexasInstrument.

[2],"TMS320C32",Ohm,1999.[3]D.Maio,D.Maltoni,R.Cappelli,J.L.Wayman,andA.K.Jain,"FVC2004:Third FingerprintVerification Competition",Proc.Int.Conf.onBiometicAuthentication(ICBA),HongKong,July2004,pp.1-7.

[4],,,,, ,ETRIJournal,2006,3.

[5]FVC2000:http://bias.csr.unibo.it/fvc2000[6]FVC2002:http://bias.csr.unibo.it/fvc2002[7]FVC2004:http://bias.csr.unibo.it/fvc2004[8] S.Prabhakar and A.K.Jain,"Decision-level Fusion inFingerprintVerification",Pattern Recognition,Vol.35,No.4,2002,pp.861-874.

[9]R.Brunelliand D.Falavigna,"Person Identification UsingMultiple Cues",IEEE Tr.Pattern Analysis and MachinePerception,Vol.17,No.10,1995,pp.955-966.

[10]L.HongandA.K.Jain,"IntegrationFacesandFingerprintsforPersoanlIdentification",Proc.3rdACCV,HongKong,Jan.1998,pp.16-23.

[11]P.Jourlinetal.,"Acostic-LabialSpeakVerification",LNCS

-94-

1206-Proc. Audio and Videobased Biometric PersonAuthentication(ABVPA),1997,pp.319-326.

[12] ,2003 ,2003.10.

[13]L.C.Jain,U.Halici,I.Hayashi,S.B.Lee,S.Tsutsui,"IntelligentBiometricTechniquesin FingerprintandFaceRecognition",CRC Press.

[14]R.BrunelliandD.Falavigna,"PersonalIdentificationUsingMultipleCues,"IEEE Trans.Pattern AnalysisandMachineIntelligence,vol.17,no.10,pp.995-966,Oct.1995.

[15]M.A.Aizerman,E.M.Braverman,and L.I.Rozonoer,"Theoreticalfoundationsofthepotentialfunctionmethodinpatternrecognitionlearning.AutomationandRemoteControl,25:821-837,1964.

[16] C.Cortes and V.Vapnik."Support vector networks",MachineLearning,20:273-297,1995.

[17] P.Dierckx,"Curve and Surface Fitting with Splines",MonographsonNumericalAnalysis",ClarendonPress,Oxford,1993.

[18]F.Girosi,"AnequivalencebetweensparseapproximationandSupportVectorMachines",A.I.Memo1606,MIT ArtificialIntelligenceLaboratory,1997.

[19] InX.Liu,P.Cohen,andM.Berthold,editors,IntelligentDataAnalysis,volume1208ofLectureNotesin ComputerScience,pages313-323,1997.

-95-

[20]N.Heckman."Thetheoryandapplicationofpenalizedleastsquaresmethodsorreproducing kernelhilbertspacesmadeeasy,1997.

[21]S.Mukherjee,E.Osuna,andF.Girosi."Nonlinearpredictionofchaotictimeseriesusingasupportvectormachine",InJ.Principe,L.Gile,N.Morgan,andE.Wilson,editors,NeuralNetworks forSignalProcessing VII- Proceedings ofthe1997IEEEWorkshop,New York,1997.IEEE.

[22]K.-R.M uller,A.Smola,G.Ratsch,B.Scholkopf,J.Kohlmorgen,and V.Vapnik.Predicting time series withsupportvectormachines.InB.Scholkopf,C.J.C.Burges,andA.J.Smola,editors,AdvancesinKernelMethods-SupportVectorLearning,pages243-254,Cambridge,MA,1999.MITPress.ShortversionappearedinICANN'97,SpringerLectureNotesinComputerScience.

[23]E.Osuna,R.Freund,andF.Girosi.Animprovedtrainingalgorithm forsupportvectormachines.InJ.Principe,L.Gile,N.Morgan,and E.Wilson,editors,NeuralNetworks forSignalProcessing VII - Proceedings of the 1997 IEEEWorkshop,pages276-285,New York,1997.IEEE.

[24]A.J.Smola."Regression estimation with supportvectorlearningmachines",Master'sthesis,TechnischeUniversitatM unchen,1996.

[25]A.J.SmolaandB.Scholkopf."Onakernel-basedmethodforpatternrecognition,regression,approximationandoperator

-96-

inversion",Algorithmica,22:211-231,1998.[26]M.O.StitsonandJ.A.E.Weston"Implementationalissuesofsupportvectormachines",TechnicalReportCSD-TR-96-18, Computational Intelligence Group, Royal Holloway,UniversityofLondon,1996.

[27]V.Vapnik,S.Golowich,and A.Smola "Supportvectormethodforfunctionapproximation,regressionestimation,andsignal processing", Advances in Neural InformationProcessingSystems9,pages281-287,Cambridge,MA,1997.MITPress.

[28]http://www.isis.ecs.soton.ac.uk/resources/svminfo/[29]SteveR.Gunn,"SupportVectorMachinesforClassificationand Regression",TechnicalReport,Faculty ofEngineering,ScienceandMathematicsSchoolofElectronicsandComputerScience,10May1998.

[30],RFID ,.[31],RFID .[32] RFID/USN,"RFID .[33],RFID,2005.[34], RFID HANDBOOK",03-00095,2004.

[35],,,RFID ,,Vol.15,No.2,2004.4.

[36],,RFID ,,Vol.15,No.2,2004.4.

-97-

[37],,,RFID/USN , 23 2,2005.2,pp.83-87.

[38],RFID/USN , 5 3,2004.9,pp.73-83.

[39],,RFID ,2005.[40],, [41]http://utenti.lycos.it/matlab[42]Arun Ross,AnilJain,"Information fusion in biometrics",PatternRecognitionLetters,Vol.24,pp.2115-2125,2003.

[43]Sung-BaeCho,JinH.Kim,"MultipleNetworkFusionUsingFuzzyLogic",IEEE Trans.onNeuralnetworks,Vol.6,No.2,1995.

[44]RichardO.duda,PeterE.Hart,DavidG.Stork,"PatternClassi-fication",SecondEdition,Wiley&Sons,Inc.,2001.

[45]Laurene V.Fausett,Fundamentals ofNeuralNetworks",PrenticeHall,l993.

[46] . 2000.[47],- ,,1999.[48]http://www.ituniv.or.kr/, , ,2004.

[49] H.C Lee and R.E.Gaensslen,Advances in FingerprintTechnology,Elsevier,1991.

[50]M.M.S.Chongetal.,"GeometricFrameworkforFingerprintImage Classification," Pattern Recognition,vol.30,no.9,pp.1475-1488,1997.

-98-

[51] N. Ratha, S. Chen, and A.K. Jain, "Adaptive FlowOrientationBasedFeatureExtractioninFingerprintImages,"PatternRecognition,vol.28,no.11,pp.1,657-1,672,1995

[52]A.K.Jain,S.Prabhakar,andS.Pankanti,"A Filterbank-based Representation for Classification and Matching ofFingerprints", International Joint Conference on NeuralNetworks (IJCNN),pp.3284-3285,Washington DC,July10-16,1999.

[53]"FingerprintClassificationandMatchingUsingaFilterbank",SalilPrabhakarA DISSERTATION Submitted toMichiganStateUniversityinpartialfulfillmentoftherequirementsforthedegreeofDOCTOROFPHILOSOPHY,ComputerScience& Engineering,2001

[54]FinalReport18-551(Spring 1999)FingerprintRecognitionGroupNumber19.

[55] Kenneth Nilsson and Josef Bigun, "Localization ofcorresponding points in fingerprints by complex filtering",Pattern Recognition Letters,24(2003)2135-2144SchoolofInformation Science,Computer and ElectricalEngineering(IDE), Halmstad University, P.O. Box 823, SE-301 18,Halmstad,Sweden.

[56]A.K.Jain,S.Prabhakar,L.Hong and S.Pankanti,"Filterbank-BasedFingerprintMatching",2000.

[57]LawrenceRabinerandBiing Hwang Juang,FundamentalsofSpeechRecognition,PrenticeHall,1993.

-99-

[58],,TMS320C32DSP (VR32) ,,vol.17,no.4pp.14-22,1998.

[59],,,1997[60],DHMM DataFusion , ,1998

[61], DSP , ,1999.

[62]www.iso.org,"ISO 14443-Part1:Physicalcharacteristics",InternationalOrganizationforStandardization,2000.

[63]www.iso.org,"ISO 14443-Part2:Radiofrequencypowerandsignalinterface",InternationalOrganizationforStandar-dization,2001.

[64]www.iso.org,"ISO 14443-Part3:Initializationandanti-collision",InternationalOrganizationforStandardization,2006.

[65]www.iso.org,"ISO 15693-Part1:Physicalcharacteristics",InternationalOrganizationforStandardization,2001.

[66]www.iso.org,"ISO 15693-Part2:Airinterfaceandinitia-lization",InternationalOrganizationforStandardization",2006.

[67]www.iso.org,"ISO 15693-Part3:Anticollisionandtrans-mission protocol",InternationalOrganization forStandardi-zation",2001.

-100-

AAA ---

, w biasb .

(fori=1,2,...,N) (A.1)

w .

(A.2)

() (primeproblem) , .

(w) w . w .

(Lagrangianmultipliers) .

(A.3)

-101-

i . J(w,b,) (saddlepoint) . w b . . w b J(w,b,) 0 .

1:

2:

(A.3) 1 2 (A.4) (A.5) . i . (A.6) nonzero .

(A.4)

(A.5)

(A.6)

-102-

BBB---MMMeeerrrccceeerrr

(3.25) Mercer . .

(eigenfunction) (eigenvalue) . (positivedefinite) .

Mercer : . .

(B.1)

i i . .

-103-

Mercer , .

- 1 , x i

.- .

Mercer ,SVM . , .

-104-

CCC---

C.1 AppendixC.1Powercircuitdiagram infingerprint

recognitionsystem

-105-

C.2 AppendixC.2Externalmemorycircuitdiagram

infingerprintrecognitionsystem

-106-

C.3 (AFS8500) -(1) AppendixC.3Fingerprintsensor(AFS8500)circuitdiagram

infingerprintrecognitionsystem -(1)

-107-

C.4 (AFS8500) -(2) AppendixC.4Fingerprintsensor(AFS8500)circuitdiagram

infingerprintrecognitionsystem -(2)

-108-

DDD---

D.1TMS320C32CPU AppendixD.1TMS320C32CPU peripheralinterface

circuitdiagram

-109-

D.2 AppendixD.2Externalmemoryinterfacecircuitdiagram

D.3 AppendixD.3SpeechCodec-chipinterface

-110-

D.4 LCD AppendixD.4LCDperipheralcircuitdiagram

-111-

EEE---RRRFFFIIIDDD

E.1 RFID AVR(CPU) AppendixE.1 AVR(CPU)modulecircuitdiagram inRFID

recognitionsystem

-112-

E.2 RFID RS-232 AppendixE.2 RS-232communicationcircuitdiagram

inRFIDrecognitionsystem

-113-

2001200120012001 , , , ,

6666 . . . .

, , , ,

. . . .

. . . .

, , , ,

, , , ,

(((()))) ....

4444

, , , ,

, , , , , , , ,

. . . .

, , , , , , , , , , , ,

, , , , , , , , , , , , , , , , , , , , , , , ,

, , , ,

, , , , , , , , , , , , , , , , , , , , , , , , , , , ,

, , , , , , , , , , , , , , , , ....

, , , , , , , , 81818181

, , , , , , , , , , , , , , , ,

, , , , , , , , ________________________________ . . . .

-114-

20202020

, , , , , , , , , , , , , , , ,

....

. . . .

, , , , , , , , , , , ,

, , , ,

. . . .

, , , , , , , , , , , , , , , , , , , , ____________________________

. . . . ________________________________________

, , , ,

....

, , , ,

. . . . , , , ,

. . . .

, , , ,

. . . .

. . . . , , , ,

. . . .

. . . .

. . . . , , , , ....

1 2 2.1 2.2 2.3 2.4 RFID

3 SVM 3.1 3.2 3.3 SVM 3.4 SVM

4 4.1 4.2 4.3 4.4 RFID

5 5.1 5.2 5.3 5.4 RFID

6 6.1 RFID 6.2 6.3 6.4 6.5 SVM 6.6 SVM

7

Documents

SSSVVVMMM을을을이이이용용용한한한다다다중중중생생생체체체인인인식식식 …repository.kmou.ac.kr/bitstream/2014.oak/8477/1/000002174450.pdf · 제5장