Classification of speech accents with neural networks

Abstract

Several neural network models including competitive learning and counter propagation are developed to identify individuals as either native or non-native speakers based on their accents. Some important speech features, such as pitch period and the first three formant frequencies, are used as inputs to the neural networks. Comparison results based on experiments are also presented. The primary contribution is that it provides a feasible approach for an assisting automatic speech recognition system in an environment in which different English accents may be used.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>