A u d io C om pr e ssion

M ultim e d ia S yste m s (M od u le 4 L e sso n 4 )

S um m ary: S ource s:
H S im ple A ud io C om pression: H D r. Z e -N ian L is
m Lo ssy : P re d iction b ase d cours e m aterial at:
http://w w w .cs.sfu.c a/C o u rse C e ntra l
H Ps ych oa coustic M o de l /3 6 5 /li/

H M PE G A ud io H M PE G A ud io:
http://w w w .m peg .org /M PE G /a ud io .h
m La ye r I a n d I I
tm l
m M P3 (M P E G L ay er II I )

Simple Audio Compression Methods

H Silence Compression - detect the "silence", similar to run-
length coding
H Adaptive Differential Pulse Code Modulation (ADPCM) e.g., in
CCITT G.721 -- 16 or 32 Kbits/sec.
m Encode the difference between two or more consecutive
signals; the difference is then quantized --> hence the loss
m Adaptive quantization
m It is necessary to predict where the waveform is headed
m Apple has proprietary scheme called ACE/MACE. A Lossy
scheme that tries to predict where wave will go in next sample.
Gives about 2:1 compression.
H Linear Predictive Coding (LPC) fits signal to speech model
and then transmits parameters of model. It sounds like a
computer talking, 2.4 kbits/sec.
H Code Excited Linear Predictor (CELP) does LPC, but also
transmits error term --> audio conferencing quality at 4.8
kbits/sec.

Psychoacoustic Model
Human hearing and voice
m Frequency range is about 20 Hz to 20 kHz, most sensitive at 1
to 5 KHz.
m Dynamic range (quietest to loudest) is about 96 dB
m Normal voice range is about 500 Hz to 2 kHz
Low frequencies are vowels and bass
High frequencies are consonants
How sensitive is human hearing?
To answer this question we look at the following concepts:
m Threshold of hearing
Describes the notion of quietness
m Frequency Masking
A component (at a particular frequency) masks components at
neighboring frequencies. Such masking may be partial.
m Temporal Masking
When two tones (samples) are played closed together in time, one can
mask the other.
Threshold of hearing
Experiment: Put a person in a quiet room. Raise level of 1 kHz
tone until just barely audible. Vary the frequency and plot

40

30
bB
20

10

0
2 4 6 8 10 12 14 16
Frequency (KHz)

H The ear is most sensitive to frequencies between 1 and 5

kHz, where we can actually hear signals below 0 dB.
H Two tones of equal power and different frequencies will not
be equally loud.
H Sensitivity decreases at low and high frequencies.

Frequency Masking
Experiment: Play 1 kHz tone (masking tone) at fixed level (60
dB). Play test tone at a different level (e.g., 1.1 kHz), and
raise level until just distinguishable. Vary the frequency of
the test tone and plot the threshold when it becomes
audible:

Frequency Masking (Contd.)

H Repeat previous experiment for various frequencies of
masking tones
Temporal Masking
H If we hear a loud sound, and then it stops, it takes a little
while until we can hear a soft tone nearby (in frequency).
H Experiment:
m Play 1 kHz masking tone at 60 dB, plus a test tone at 1.1 kHz at
40 dB. Test tone can't be heard (it's masked).
m Stop masking tone, then stop test tone after a short delay.
m Adjust delay time to the shortest time when test tone can be
heard (e.g., 5 ms).
m Repeat with different level of the test tone and plot:

Net effect of masking:

MPEG Audio
Facts
H The two most common advanced (beyond simple ADPCM)
techniques for audio coding are:
m Sub-Band Coding (SBC) based
m Adaptive Transform Coding based
H MPEG audio coding is comprised of three independent layers.
Each layer is a self-contained SBC coder with its own time-
frequency mapping, psychoacoustic model, and quantizer.
m Layer I: Uses sub-band coding
m Layer II: Uses sub-band coding (longer frames, more
compression)
m Layer III: Uses both sub-band coding and transform coding.
H MPEG-1 Audio is intended to take a PCM audio signal sampled
at a rate of 32, 44.1 or 48 kHz, and encode it at a bit rate
of 32 to 192 kbps per audio channel (depending on layer).
More Facts
H MPEG-1: Bitrate of 1.5 Mbits/sec for audio and video About
1.2 Mbits/sec for video, 0.3 Mbits/sec for audio
m (Uncompressed CD audio is 44,100 samples/sec * 16 bits/sample
* 2 channels > 1.4 Mbits/sec)
H Compression factor ranging from 2.7 to 24.
H With Compression rate 6:1 (16 bits stereo sampled at 48
KHz is reduced to 256 kbits/sec)
m Under optimal listening conditions, expert listeners could not
distinguish between coded and original audio clips.
H Supports one or two audio channels in one of the four modes:
1. Monophonic -- single audio channel
2. Dual-monophonic -- two independent channels, e.g., English and
French
3. Stereo -- for stereo channels that share bits, but not using
Joint-stereo coding
4. Joint-stereo -- takes advantage of the correlations between
stereo channels

MPEG Coding Algorithm

Input Filter into Output
Allocate bits Format
Critical Bands
(Quantization) BitStream
(Sub-band filtering

Compute
Masking
(Psychoacoustic
Model)

1. Use convolution filters to divide the audio signal (e.g., 48 kHz

sound) into 32 frequency sub-bands. (sub-band filtering)
2. Determine amount of masking for each band caused by nearby
band using the psychoacoustic model .
3. If the power in a band is below the masking threshold, don't
encode it.
4. Otherwise, determine number of bits needed to represent the
coefficient such that, the noise introduced by quantization is
below the masking effect (Recall that one fewer bit of
quantization introduces about 6 dB of noise).
5. Format bitstream

Masking and Quantization (Example)

H Say, performing the sub-band filtering step on the input
results in the following values (for demonstration, we are
only looking at the first 16 of the 32 bands):

Band 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

Level 0 8 12 10 6 2 10 60 35 20 15 2 3 5 3 1

H The 60dB level of the 8th band gives a masking of 12 dB in

the 7th band, 15dB in the 9th. (according to the
Psychoacoustic model)
H The level in 7th band is 10 dB ( < 12 dB ), so ignore it.
H The level in 9th band is 35 dB ( > 15 dB ), so send it.
H We only send the amount above the masking level
H Therefore, instead of using 6 bits to encode it, we can use 4
bits -- a saving of 2 bits (= 12 dB).
H determine number of bits needed to represent the coefficient
such that, the noise introduced by quantization is below the
masking effect [noise introduced = 12bB; masking = 15 dB]
MPEG Coding Specifics

12 12 12
samples samples samples

Sub-band filter 0

Sub-band filter 1
Audio
Samples
Sub-band filter 2
. .
. .
. . .
.
12 12 12
.
samples samples samples

Sub-band filter 31
Layer I
Frame
Layer II, III
Frame

MPEG Coding Specifics

H MPEG Layer I
m Filter is applied one frame (12x32 = 384 samples) at a time. At 48 kHz,
each frame carries 8ms of sound.
m Uses a 512-point FFT to get detailed spectral information about the
signal. (sub-band filter). Uses equal frequency spread per band.
m Psychoacoustic model only uses frequency masking.
m Typical applications: Digital recording on tapes, hard disks, or magneto-
optical disks, which can tolerate the high bit rate.
m Highest quality is achieved with a bit rate of 384k bps.
H MPEG Layer II
m Use three frames in filter (before, current, next, a total of 1152
samples). At 48 kHz, each frame carries 24 ms of sound.
m Models a little bit of the temporal masking.
m Uses a 1024-point FFT for greater frequency resolution. Uses equal
frequency spread per band.
m Highest quality is achieved with a bit rate of 256k bps.
m Typical applications: Audio Broadcasting, Television, Consumer and
Professional Recording, and Multimedia.
MPEG Coding Specifics
H MPEG Layer III
m Better critical band filter is used
m Uses non-equal frequency bands
m Psychoacoustic model includes temporal masking effects, takes
into account stereo redundancy, and uses Huffman coder.
Stereo Redundancy Coding:
m Intensity stereo coding -- at upper-frequency sub-bands,
encode summed signals instead of independent signals from left
and right channels.
m Middle/Side (MS) stereo coding -- encode middle (sum of left
and right) and side (difference of left and right) channels.

Effectiveness of MPEG Audio

Layer Target Ratio Quality* Quality at
bit-rate at 128 kbps
64 kbps
Layer I 192 kbps 4:1 -- --

Layer II 128 kbps 6:1 2.1 to 2.6 4+

Layer III 64 kbps 12:1 3.6 to 3.8 4+

*Quality factor:
m 5 perfect
m 4 - just noticeable
m 3 - slightly annoying
m 2 annoying
m 1 - very annoying