The long term objective of this research in Phase I and Phase II is to develop a speaker-independent, voice-activated telephone interface for use by quadriplegics and others, with the innovative capability of handling speech produced by people who require ventilators. Specific objectives include: (1) collecting and analyzing """"""""ventilator speech"""""""" produced by people with spinal cord injuries at different phases of rehabilitation; (2) modifying Emerson & Stern's existing SOLILOQUY R Voice Interface Software to handle breath-related interference and ignore non-speech interjections, without compromising SOLILOQUY's current speaker-independence; and (3) determining if an adequate telephone interface can be affixed to a halo orthotic. These objectives will be met by (1) recording telephone numbers spoken by patients at Sharp Hospital's Rehabilitation Center; (2) using a series of in-house software tools for speech analysis; (3) iteratively improving SOLILOQUY's microphone recognition algorithms and testing the results; (4) making similar changes to telephone-based algorithms; and (5) testing a cervical orthosis-mounted phone interface. Other voice-activated phones are speaker-dependent, requiring users to """"""""re- train"""""""" the system every time their voice changes. No other system has SOLILOQUY's linguistic ability to ignore a non-speech sound by treating it as extraneous speech.
