Neural network models for recognition of consonant-vowel ((CV)-V-n) utterances

In this paper, we present an approach based on neural network models for recognition of utterances of syllable-like units in Indian languages. The distribution capturing ability of an autoassociative neural network model is exploited to perform nonlinear principal component analysis for compressing the size of the feature vector. A constraint satisfaction model is proposed to incorporate the acoustic-phonetic knowledge and to combine the outputs of subnets to arrive at the overall decision on the class of an input utterance.