On the Design of Sub-band Coders for Low-Bit-Rate Speech Communication
01 May 1977
In recent work by Crochiere, Webber and Flanagan, 1 an approach to speech encoding has been proposed which is based on the partitioning of the speech band into sub-bands and encoding the sub-bands individually. The technique offers attractive possibilities for coding speech economically at bit rates in the range of 7.2 to 16 kb/s. At 16 kb/s good quality encoding, comparable to that of 26.5 kb/s adaptive differential (fixed predictor) PCM (ADPCM) encoding, is possible. Potential applications exist in areas of narrow-band communications, mobile radio, and voice storage applications. When the bit rate is extended down into the upper data rate range of 9.6 and 7.2 kb/s, moderate quality encoding can be achieved comparable to that of 19 and 18 kb/s adaptive delta modulation (ADM), respectively. Interesting potential applications exist for voice coordination on digital data lines and for secure voice communications by digital encryption and transmission over conventional data lines. In the design of sub-band coders, a variety of issues and "trade-offs" must be dealt with. The number of sub-bands, the partitioning of sub747