Finite State Network, Isolated Word, Real Time Automatic Speech Recognizer based on DSP32C.

This work describes a Finite State Network (FSN), isolated word, speaker dependent, real time Automatic Speech Recognition (ASR) system. FSNs are used to represent grammars belonging to regular language class [ Aho et al., 1974 ] . It is well known that regular languages are powerful enough to generate useful subsets of the natural language. On the other hand, they can be easily integrated into an ASR system [ Bahl et al., 1983 ] [ Rabiner and Juang, Jan 1988 ] embedding grammar constraints to reduce the search space, hence the time spent in recognition, and increase the recognition rate [ Rabiner and Levinson, 1981 ] [ Kai-Fu Lee, April 1988 ] . For isolated word, speaker dependent ASR systems, a template matching recognition algorithm based on Dynamic Time Warping (DTW) [ Velichko and Zagoruyko, 1970 ] [ Vintsyuk, 1968 ] [ Silverman and Morgan, July 1990 ] ensures a high recognition rate and quick development. A single AT&T LSI DSP32C board plugged into an ISA bus has been used as co..