A self-organizing neural network approach for the acquisition of phonetic categories

Abstract

We present a neural network approach to the process of acquisition of phonetic categories in infants. In our approach we investigate the question to what extend the development of phonetic categories can be described by a self-organizing process. Simulation results show that with digitized speech as input, the network is able to learn representations of the vowel categories in the input set.