Adaptive Quantization in Differential PCM Coding of Speech

The advantages of coding speech digitally are well known.1 Expected benefits include low costs per line, ease of maintenance, and highquality signal regeneration at repeaters. Furthermore, digital coding is well matched to current technology in terms of readily available integrated circuit hardware. Results from speech-coding research are now beginning to specify techniques that are nearly optimal for a given bit rate, a given channel quality, and a given degree of coder complexity. Finally, the subject of direct digital conversion between alternative code formats is being widely studied, and simple techniques have already been proposed for some specific conversions. The coder discussed in this paper is believed to be efficient and robust for speech coding at bit rates of 24 to 32 kilobits/second (kb/s). Other refinements of differential PCM (DPCM) 2 are based, at least in part, on adaptive prediction. 3-6 These techniques offer considerable potential for bandwidth compression,3 but are typically hard to 1105