Predictive Kohonen map for speech features extraction

Abstract

Some well known theoretical results concerning the universal approximation property of MLP neural networks with one hidden layer have shown that for any function / from [0,1]n to R, only the output layer weights depend on f. We use this result to propose a network architecture called the predictive Kohonen map allowing to design a new speech features extractor. We give experimental results of this approach on a phonemes recognition task. © Springer-Verlag Berlin Heidelberg 2005.