Tracking the pseudo-pitch of unvoiced sounds: a hand-free interface modality for disabled users

We developed a prototype of interface for persons with severe motor disabilities based on a new entry modality, namely unvoiced sounds like hissing, inspiration or expiration. Unlike sip-and-puff sensors, this modality uses the height of the sound as an analog parameter. Unvoiced sounds have no fundamental frequency (pitch). However they present a formant (pseudo-pitch), i.e., a peak of energy (typically above 800 Hz) that is perceived as the height of the sound. The pseudo-pitch can be partially controlled even in case of speech impairment. In our prototype, the pseudo-pitch is tracked in real time. Commands are entered by stabilizing the pseudo-pitch at the desired frequency. A voice-detection algorithm shuts off the entry during speech and strong noises, e.g., cough. It is possible to enter commands at a low volume and to speak aloud in the same microphone. Preliminary tests indicate that with the help of visual feedback, it is possible to produce 4 different commands and attain throughput above that throughputs above 166 bits/minute can be attained