deepsh.it - networking

Voice Codecs

2 groups of voice codecs:

1. Waveform codecs
    Exploit the redundant characteristics of the waveform itself.
    ex:
    Pulse Code Modulation PCM
    Adaptive Differential Pulse Code Modulation ADPCM

2. Source codecs
    Newer compression techniques that exploit knowledge of the source characteristics of speech generation
    ex:
    Linear Predictive Coding (PLC)
    Code Exited Linear Prediction (CELP)
    Multipulse, Multilevel Quantization (MP-MLQ)

Pulse Code Modulation (PCM):
The most common method of encoding an analog voice signal.
It’s based on Nyquist theorem (if you sample at twice the highest frequency, you’ll achieve good-quality voice).
Encode using 8 bits samples to encode the amplitude of speech.
Will guarantee good quality voice up to 4000 Hz by sampling at 8000Hz
PCM use 2x4000=8000 Hz sampling. (4000 Hz sufficient for voice quality)
The transmission rate is 8000 x 8 bit = 64000 kbps.
2 basic PCM variations: u-law & a-law, u-law has a slight lower S/N compare to a-law.

Adaptive Differential Pulse Code Modulation (ADPCM):
Encodes using 4 bits samples.
Transmission rate: 32 kbps.
Unlike PCM, the 4 bits do not directly encode the amplitude of the speech, but do encode the differences in amplitude, as well as the rate of change of that amplitude.

Voice Coding Standard:

ITU-T standardizes PCM, ADPCM, CELP, MP-MLQ PCM coding schemes in its G-series recommendation:

G.711:
64 Kbps PCM voice coding technique.
used in the public phone network and through PBX

G.726:
ADPCM encoding at 40,32, 24 and 16 Kbps.
Some public phone network and PBX support it.

G.728:
low delay variation of CELP voice compression at 16 Kbps

G.729:
CELP at 8 Kbps

G.723.1
compression at low bit rate

Compression Method	Bit Rate (Kbps)	Sample Size (ms)	MOS Score
G.711 PCM (u-law, a-law)	64	0.125	4.1
G.726 ADPCM	32	0.125	3.85
G.728 LD-CELP	15	0.625	3.61
G.729 CS-ACELP	8	10	3.92
G.729a CS-ACELP	8	10	3.7
G.723.1 MP-MLQ	6.3	30	3.9
G.723.1 ACELP	5.3	30	3.65

LD-CELP: Low Delay-CELP
CS-ACELP: Conjugated Structure Algebraic CELP
MOS: mean opinion score

$Date: Wed Jan 12 18:00:48 CET 2005 $