The goal of the proposed research is to determine the feasibility of developing a speaker-independent speech recognition I system that can be implemented on currently available hardware components. The svsteiii uses J'eature-based approach to speech recognition, in which the perceptually relevant features of speech are extracted from the signal and used to identifv speaker-independent patterns for the consonant and vowel segments in an utterance. Basic research performed by the principal investigator at Carnegie-Mellon University demonstrates that the feature-based approach is more powerful than the template matching approach used in all commercially available systems. In order to determine the feasibility of a real time system, it is necessary to develop a prototype system and then examine the computational requirements of the algorithms in light of currently available hardware components. This proposal describes the software configuration of the proposed system and the research that is needed to complete software development and determine the computational requirements of the resulting algorithms.The potential commercial application as described by the awardee: The proposed research may lead to the development of voice data entry terminals, voice control of machinery and processes, and voice operated typewriters (automatic dictation).
