A direct voice input man-machine interface strategy to provide voice access for severely impaired speakers

A description is given of the initial stages in the development of a simple time-encoded digital speech (TES) isolated word recognition, (IWR) system, incorporating automatic token set selection (ATSS) procedures as a means of providing an effective man-machine interface for severely handicapped speakers. Three aspects of the investigation are emphasised. The use of automatic diagnostic routines, as a means of assessing and selecting acoustic utterances produced by severely impaired speakers is discussed. The characteristics of so called sentence generator algorithms, developed to capitalise on the advantageous features of the TES coding format and ATSS are described. Finally, the first results of exercising a semi-portable test bed embodiment against a small number of cooperative but severely impaired speakers are presented. A working first prototype is demonstrated