Layered neural networks applied to the recognition of voiceless unaspirated stops

The behavior of layered neural nets when they are applied to speech recognition is studied. The input to the neural net is a feature vector which describes the characteristics of the voiceless unaspirated stops. The function of the neural net is to classify the articulation place of the stop consonants. In this study, the neural network classifier is compared with the Bayes classifier to reveal the advantages gained by using a neural network as a classifier. The effect of the number of the processing units in the hidden layer is examined. A method of minimizing the degradation in performance of an existing neural net when one of the hidden processing units misses is proposed