Research in continuous speech recognition

This final report describes the results of a three-year project in an ongoing basic research effort to develop techniques for the automatic recognition of speech sounds (phonemes) in unrestricted continuous speech. A basic working phonetic recognition system has been designed, implemented on a VAX 11/780, and tested. At over 75% phonetic accuracy, the results have already exceeded previously-published phonetic recognition results on continuous speech for a single speaker. Our future goal will be to increase the recognition accuracy to over 80% for many speakers.