View
225
Download
0
Category
Preview:
Citation preview
8/17/2019 Adsp 03 Ac Intro Ec623 Adsp
1/15
Audio Coding
Introduction
S. R. M. Prasanna
Dept of ECE,
IIT Guwahati,
prasanna@iitg.ernet.in
Audio Codin – . 1/
w w w . j n t u w or l d . c om
http://prosper.sourceforge.net/http://prosper.sourceforge.net/
8/17/2019 Adsp 03 Ac Intro Ec623 Adsp
2/15
Goal of Audio Coding
Terms Coding and Compression are usedinterchangeably.
Goal of audio coding is to develop methods for compactdigital representation of audio signals.
Efficient transmission or storage.
Minimum number of bits with transparent perceptualquality.
Audio Codin – . 2/
w w w . j n t u w or l d . c om
8/17/2019 Adsp 03 Ac Intro Ec623 Adsp
3/15
First Generation Audio Coders
Digital representation of audio signals.
Compact Disc (CD) is the digital storage medium.
Sampling frequency is 44.1 kHz and Bit rate 16bits/sample
20 kHz audio spectrum + 2.05 guard band = 22.05kHz
Sampling freq = 22.05× 2 = 44.1kHz .
Data rate:44100× 16 = 705.6 kb/s for mono705.6× 2 = 1.41 Mb/s for stereo
Audio Codin – . 3/
w w w . j n t u w or l d . c om
w
8/17/2019 Adsp 03 Ac Intro Ec623 Adsp
4/15
Second Generation Audio Coders
For network and wireless multimedia digital audio.
Bandwidth is the severe constraint.
At the same time, end-users need CD quality.Conflicting requirements.
Goal is to reduce data rate without compromising on
the perceptual quality.Led to several audio compression algorithms.
Exploit both perceptual irrelevancies and statistical
redundancies.
Audio Codin – . 4/
w w w . j n t u w or l d . c om
w
8/17/2019 Adsp 03 Ac Intro Ec623 Adsp
5/15
Third Generation Audio Coders
Lossless audio
Spatial audio
Real-time source localizationHead related transfer function (HRTF)
Immersive audio
Audio Codin – . 5/
w w w . j n t u w or l d . c om
w
8/17/2019 Adsp 03 Ac Intro Ec623 Adsp
6/15
Audio Coding Methods
PCM (1.41 Mb/s).
DPCM (0.75 x PCM data rate).
ADPCM (0.5 x PCM data rate).Not much data rate reduction.
Need for high compression methods driven by potential
applications.New approaches for audio coding based on theprinciples of psychoacoustics.
Audio Codin – . 6/
w w w . j n t u w or l d . c om
w
8/17/2019 Adsp 03 Ac Intro Ec623 Adsp
7/15
Psychoacoustics
Characterizing human auditory perception.
Time-frequency analysis capabilities of the inner ear.
Perceptually irrelevant audio signal information.Contributions from psychoacoustics:
Perceptual entropy
Auditory filter bankPerceptual entropy deals with estimate of thefundamental limit of transparent audio signal
compression.Auditory filter bank based on the time-frequencyanalysis capabilities of the inner ear.
Audio Codin – . 7/
w w w . j n t u w or l
d . c om
w
8/17/2019 Adsp 03 Ac Intro Ec623 Adsp
8/15
Some Audio Coding Standards
MPEG-1 Audio (1992).
MPEG-2 Audio (1996).
MPEG-4 Audio v1 (1999).MPEG-4 Audio v2 (2000)
Audio Codin – . 8/
w w w . j n t u w or l
d . c om
w
8/17/2019 Adsp 03 Ac Intro Ec623 Adsp
9/15
Block Diagram of Generic Audio Coder
Audio Codin – . 9/
w w w . j n t u w or l
d . c om
w w
8/17/2019 Adsp 03 Ac Intro Ec623 Adsp
10/15
Principle of Generic Audio Coder
Segment input signals into quasi-stationary frames of2-50 ms.
Time-frequency analysis estimates the temporal andspectral components of each frame.
TFA approach employed is based on human auditorysystem.
Objective is to extract a set of time-frequencyparameters that are robust to quantization according toa perceptual distortion metric.
Perceptual distortion control is achieved by apsychoacoustic signal analysis section that estimatessignal masking power based on psychoacoustic
principles.
Audio Codin – . 10/
ww w . j n t u w or l d . c om
w w
8/17/2019 Adsp 03 Ac Intro Ec623 Adsp
11/15
Principle of AC (contd.)
Psychoacoustic model delivers masking thresholds thatquantify the maximum amount of distortion at eachpoint in the time-frequency plane such that quantization
of the time-frequency parameters does not introduceaudible artifacts.
Psychoacoustic model allows the quantization section
to exploit perceptual irrelevancies.Final redundancy removal based on the perceptualentropy coding scheme.
Audio Codin – . 11/
ww w . j n t u w or l d . c om
w w
8/17/2019 Adsp 03 Ac Intro Ec623 Adsp
12/15
Audio Coder Attributes
Audio reproduction quality.
Operating bit rates.
Computational complexity.Codec delay.
Channel error robustness.
High quality audio at low bit rates (
8/17/2019 Adsp 03 Ac Intro Ec623 Adsp
13/15
Types of Audio Coders
Based on the signal model or analysis-synthesistechnique.
LP
Transform
Subband
Sinusoidal
Audio Codin – . 13/
w w . j n t u w or l d . c om
w w
8/17/2019 Adsp 03 Ac Intro Ec623 Adsp
14/15
AC-Expt.1
Effect of Sampling Frequency and Bit Resolution
Objective is to analyze the effect of sampling frequency
and bit resolution on the perceptual quality of audio.Take a CD quality music signal of 1 sec, sampled at44.1 kHz with 16 bits/sample and perform the following.
Change its sampling frequency to 16, 8 and 4 kHz.Keep bit resolution constant at 16 bits/sample.Consider about 50 ms segment in a high energyregion.
Plot the time domain and DFT spectra for all thefour cases.Comment on the effect of different sampling
frequency.Comment also on the perceptual quality of theaudio.
Audio Codin – . 14/
w w . j n t u w or l d . c om
AC E 1 w w
8/17/2019 Adsp 03 Ac Intro Ec623 Adsp
15/15
AC-Expt.1
Effect of Sampling Frequency and Bit Resolution
Change its bit resolution to 8, 4 and 1 bits/sample.
Keep sampling frequency constant at 44.1 kHz.Consider the same 50 ms segment in a high energyregion.
Plot the time domain and DFT spectra for all the
four cases.Comment on the effect of different bit resolutions.Comment also on the perceptual quality of the
audio.
Audio Codin – . 15/
w w . j n t u w or l d . c om
Recommended