On Predictive Quantizing Schemes

Predictive quantizing schemes employ prediction to exploit the inherent redundancy of input signals; the difference between the actual sample of an input signal and its estimate is quantized and transmitted to the receiver in a digital format. If the samples of the input are highly correlated, then the variance of the samples of the difference signal will be significantly less than the variance of the input samples. Hence, the overall error between input and output of the communication system will be lower than that of a conventional pulse code modulation (PCM) system. The first part of this paper compares PCM and two differential (predictive) pulse code modulation systems (see Fig. 1). A noise-feedback coding structure is then used as a framework of a unified analysis of predictive quantizing schemes (including those of Fig. 1) on the basis of a frequency-weighted error criterion. The last part of this paper ex1499 (a) PCM CHANNEL Fig. 1--Structures of PCM, DPCM, and D*PCM coders.