Discriminative coding with predictive neural networks

We present a speech coding called NPC (neural predictive coding) based on an original neural predictive structure. The system is designed to predict the samples of a speech signal window from previous ones. The goal of this coding is to extract the signal discriminative features relative to the database from which it is extracted. After a precise description of NPC coding, we discuss the first phase of the algorithm: the adjustment of the parameters of the coder. We present experiences which show an improvement on telephone speech recognition compared to LPC. The signal windows are extracted from the telephone speech signal database NTIMIT