What makes voice compression possible is the power of Digital Signal Processors. DSPs have continued to increase in performance and decrease in price over time, and as they have, it has made it possible to use new compression schemes that offer better quality and use less bandwidth. The power of the DSP makes it possible to combine this traffic onto a line that formerly supported perhaps only a LAN connection, but now can support voice, data, and LAN integration.
Looking at this chart, quality and bandwidth tend to trade off. PCM is the standard 64Kbps scheme for coding voice; it is the standard for toll quality. The other compression schemes – ADPCM at 32Kbps, 24Kbps and 16Kbps – offer less quality but more bandwidth efficiency. The newer compression schemes -LDCELP at 16Kbps and CS-ACELP at 8Kbps – offer even higher efficiency but with very high quality very acceptable in a business environment.
ADPCM—Adaptive Differential Pulse Code Modulation: consumes only 32 Kbps compared to the 64 Kbps of a traditional voice call; often used on long-distance connections.
LPC—Linear predictive code: a second group of standards that provide better voice compression and, at the same time, better quality. In these standards, the voice coding uses a special algorithm, called linear predictive code (LPC), that models the way human speech actually works. Because LPC can take advantage of an understanding of the speech process, it can be much more efficient without sacrificing voice quality.
CELP—Code-Excited Linear Predictive voice compression: uses additional knowledge of speech to improve quality.
CS ACELP: Further improvements known as conjugate structure algebraic compression enable voice to be coded into 8-kbps streams. There are two forms of this standard, both providing speech quality as good as that of 32-kbps ADPCM.
Voice Quality Guidelines
Silence Suppression by Voice Activity Detection
Voice activity detection (VAD) provides for additional savings beyond that achieved by voice compression.
Telephone conversations are half duplex by nature , because we listen and pause between sentences. Sixty percent of a 64-kbps voice channel typically contains silence. VAD enables traffic from other voice channels or data circuits to make use of this silence.
The benefits of VAD increase with the addition of more channels, because the statistical probability of silence increases with the number of voice conversations being combined.