Multistage self-excited linear predictive speech coder

Presents an approach to low-rate speech coding based on multistage self-excited linear prediction (SELP). The coder has much lower complexity than a code-excited linear prediction (CELP) coder, but yields comparable output speech quality at the same transmission rate. In the proposed multistage SELP coder, the long-term predictor is modified to model better the onset of a voiced section of speech or voice with high pitch.