NEURAL NETWORKS FOR DISCRIMINATION AND MODELIZATION OF SPEAKERS

This article reviews current research on neural network systems for sp eaker recognition tasks. We consider two main approaches, the first on e relies on direct classification and the second on speaker modelizati on. The potential of connectionist models for speaker recognition is f irst presented and the main models are briefly introduced. We then pre sent different systems which have been recently proposed for speaker r ecognition tasks. We discuss their respective performances and potenti als and compare these techniques to more conventional methods like vec tor quantization and Hidden Markov models. The paper ends with a summa ry and suggestions for further developments.