SGER: Exploration of a Neurological Model to Improve the Extraction of Linguistic Features in Speech

University Of Maryland, College Park, College Park MD

Investigators

Abstract

With National Science Foundation support, Dr. Carol Espy-Wilson will study a knowledge-based speech recognition system's performance in noise. The research is linguistically motivated way, in that it seeks acoustic correlates of linguistic features. This project emphasizes two components of the PI's larger research agenda: (1) analysis of signal representation to ensure its robustness, and (2) application of a neural model to enhance the signal-to-noise ratio before or during the extraction of a knowledge-based speech signal representation. The broader impacts of the project lie in its progress towards speaker-independent speech recognition, which has practical applications in industry, education and speech rehabilitation.