A long-term predictive ADPCM coder with short-term prediction and vector quantization

A speech coder structure based on the long-term predictive ADPCM (adaptive differential pulse-code modulation) coder with vector quantization of pitch coefficients, the LPC (linear predictive coding) coefficients, and the residual signal is presented. In order to improve the robustness against variations in the average spectral envelope a backward adaptive microphone and speaker adaptation filter is used for preprocessing the speech. With this coder, a good delay/quality tradeoff can be obtained. Simulations of a 8.133 kb/s coder with an algorithmic delay of only 7.5 ms show that a very good speech quality is produced. >