Acoustic-Phonetic Knowledge and Speech Recognition

University Of Maryland, College Park, College Park MD

Investigators

Abstract

With National Science Foundation support, Dr. Carol Espy-Wilson will conduct three years of interdisciplinary research. This project combines knowledge and techniques from engineering, speech science, and computational linguistics to develop a robust automatic speech recognition system that works well for a large and diverse population. This project includes the development of (a) acoustic parameters that target the linguistic information in the speech signal, and (b) a new paradigm for recognition that integrates current speech knowledge with a powerful statistical framework. The knowledge-based speech recognition system serves as a model of the early stages of speech perception. Its development is a tool for understanding and coping with variation in speech that occurs within and across speakers due to differences in gender, age and emotional states. New insights will be gained in the areas of speech production, speech perception, and acoustic phonetics. Further, different components of the system can serve as speech analysis tools to be used in various applications including speech coding, automatic speech transcription, speech enhancement, and speech training aids for speech pathologists and their clients.