Voiceband data compression at 16 kb/s

The effects of waveform coding on voiceband data are explored. In particular, voiceband data compression at 16 kb/s is considered, and the performance and attributes of several 16 kb/s compression algorithms (APC, ATC, CELPC, PCM, SBC, and VQ) are discussed. Three mechanisms for improving performance are noted: decreasing the error variance, increasing the correlation between signal and error, and limiting the maximum excursion of the error. It is shown that PCM (pulse code modulation) successfully capitalizes on the latter two mechanisms to deliver performance comparable or superior to much more complicated coders. It is shown that good performance for speech does not imply good performance for data and that coders optimized for voiceband data significantly outperform their speech-specific counterparts.<
>