Segment-Holistic Speech Recognition System Based on Model of Brain Hemispheric Interaction in Speech Perception

Abstract

This paper is devoted to the development of two-channel segment-holistic system of medium-vocabulary speech recognition. Theoretical basis of the given system is the concept about functional asymmetry of cerebral hemispheres in human speech perception. The structure of automatic speech recognition system based on interaction of segment (phoneme-by-phoneme) and holistic (word-by-word) recognition subsystems is proposed. The segment subsystem is realized by means of neural networks and dynamic programming combination, and in the holistic subsystem the fuzzy matching of speech patterns is used.