Speech input for persons with speech impairments

Abstract

Some people have impairments which made their speech quite difficult for other people to understand. However, it may be that computers could be trained to reliably recognize their speech. This paper will present the results of a preliminary, small-scale experiment which produced encouraging results to support this proposition.

This project used the technique of Time Encoded Speech and was tested using a person with cerebral palsy. It showed a sufficiently high level of recognition to suggest that the technique could be applied, though it might be restricted to applications involving a limited vocabulary.