Fully vector-quantized neural network-based code-excited nonlinear predictive speech coding

I Recent studies have shown that non-linear prediction can be implemented with neural networks, and non-linear predictors will on average achieve about 2 3 improvement in prediction gain over conventional linear predictors. In this paper, we take the advantage of non-linear prediction with neural network, apply it to predictive speech coding and attempt to improve the speech coding performance. Our studies are concentrated on non-linear prediction with neural network, non-linear predictive vector quantiser, non-linear long term (pitch) speech prediction, the output variance of the non-linear predictive synthesis lter to its input distortion and fully vector quantised code-excited non-linear predictive speech coding. The above studies have resulted in a fully vector quantised CENN-excited non-linear predictive speech coder. Performance evaluations and comparisons with the linear predictive speech coding are made in this paper.