Title: Non-Linear Speech coding with MLP, RBF and Elman based prediction

Abstract: In this paper we propose a nonlinear scalar predictor based on a combination of Multi Layer Perceptron, Radial Basis Functions and Elman networks. This system is applied to speech coding in an ADPCM backward scheme. The combination of this predictors improves the results of one predictor alone. A comparative study of this three neural networks for speech prediction is also presented.