Prediction techniques applied to Differential Pulse-Code Modulation systems for encoding speech signals

Differential pulse-code modulation (DPCM) is an efficient digitization technique for encoding speech signals. The two principal components of a DPCM system are the quantizer and the predictor, either or both of which can be adaptive. This thesis describes the investigation of various differential pulse-code modulation systems. Initially, fixed, i.e. time-invariant, predictors using long-term signal statistics of the speech signal are examined. The performance of such a predictor in a DPCM system having a fixed quantizer is studied. Then by replacing the fixed predictor with one whose coefficients are calculated at fixed time intervals, the performance of the encoder, in terms of signal to noise ratio (SNR), is improved by 3-5 dB. A further improvement of 2–3 dB in SNR is obtained when an adaptive quantizer is used in the DPCM system. However, the block adaptive predictor requires the transmission of prediction coefficients to the receiving end at the expense of an increase in the channel band width. [Continues.]