A novel adaptive predictor DPCM system for speech signals

A novel Adaptive Predictor Differential Pulse Code Modulation (DPCM) technique has been developed in which the predictor, rather than the quantizer, is adapted. The z-plane pole locations are adjusted on a sample-by-sample basis, tuning the predictor to track the instantaneous frequency of the input speech waveform. The adaptive algorithm employs a simple estimate of the correlation coefficient of the input signal, based upon the three previous samples. The system has been proven operable at bit rates as low as 8 kbs, resulting in improved performance over more conventional DPCM systems. Robust performance has been obtained with bauded data signals as well as speech signals with input SNRs as low as 0 dB, and under conditions of crosstalk as high as 0 dB relative to input speech.