An Adaptive PCM System Designed for Noisy Channels and Digital Implementations

We propose a new scheme for adaptive quantization which is particularly well suited to the digital implementation of PCM and DPCM structures. In the course of this work, we have developed arithmetics for the digital processing that are useful as well in the implementation of previously existing schemes for robust quantization. The exacting requirements on adaptive quantization stemming from the broad dynamic range and rapid transient behavior of speech are well known. Two additional objectives are given equal importance here: (i) To make the system robust in the presence of channel errors. Thus, while channel errors may cause the quantizer adaptations in transmitter and receiver to be put out of synchronization,* a mechanism must exist which acts to rapidly restore the synchronization during periods of error-free transmission, (ii) To minimize the complexity of the digital realization; specifically, to minimize the length of the internal words in the digital processors and to facilitate the multiplexing of the hardware. Systems do exist in the literature for robust quantization in the presence of noisy channels; one such system is described below in some detail. However, the problem of synchronizing the quantizer adaptations in the transmitter and receiver in digital implementations, where the constraint of finite precision arithmetic exists, has not been addressed previously. We prove that, to provide the necessary synchronization capability without impairing the quality of speech reproduction, it is necessary to use an unexpectedly large number of bits in the internal words of the digital processors at both sites and, additionally, to carefully specify the internal arithmetic (which we do).