New compression and decompression of speech signals by a Neural Predictive Coding (NPC)

In this paper, we present a new method of speech compression and decompression based on a Neural Predictive Coding of speech signals. The NPC system is designed to predict the samples of a speech signal window from previous ones. In the coder/decoder that we proposed the transmitted data is computed from the prediction error of the NPC (difference between the sample and its corresponding prediction calculated by the NPC ). The initial goal of the NPC is to extract the signal discriminative features relative to the database which it is extracted. After a precise description of NPC coding, we discuss about the first phase of the algorithm: the adjustment of the parameters of the coder. Then we explain the compression and decompression algorithms. To finish, we present an example and some results on this technic of compression. Key-Words: Speech compression, Speech decompression, Neural Networks, Speech Coding, Speech Prediction.