A palm-sized keying unit has been designed w-id built by us which responds to combined thumb and finger motions. In the prototype now under development, this handheld unit (9x7xl.5cm) Itself contains a powerful 64180 microprocessor which, In addition to reading and decoding thumb-and finger-position. Input can run at clock speeds (8 KHz) compatible with the recreation of high-quality speech from digital recordings of normal speakers. in the proposed research, 1) a 12-blt high-speed digital-to-analog converter and audio output stages will be Integrated Into the keying unit computer, 2) samples of a functional English phoneme set will be digitized and analyzed an laboratory equipment, and 3) 64180 machine language program will be devised to output these samples sequentially to form unlimited-vocabulary speech. Finally, 4) the programs and sampled phonemes will be written to a high-density EPROM chip for installation in the portable keying unit, so that this speech can be generated in real time in response to use commands. Anticipated Result It is estimated that over 1/2 million people in the United States are unable to speak. Until now,, no portable device has existed which might allow these persons to speak In a near-normal fashion, and permit them to attain far greater educational, social and vocational levels than with the bulky,, robotic sounding units now available. In the present research, achieving this goal was deemed to require a) a man-machine interface to specify speech and b) a synthesizer to perform the specified speech as sound both of which should be Inconspicuously carried and operated. We are developing a handheld keying unit, operated by combined motions of the finger and thumb,, which may provide facile Input for persons with Intact coordination of at least one hand. To provide realistic speech, we hope to develop the technique of digital sampling, in which the keying unit's computer would play back short segments of prerecorded human voices In sequences to farm words and sentence. Phase I will entail digitally recording human voice sounds, preparing them for output, and writing programs to output the sounds sequentially. Phase II will Involve developing prototypes of more sophisticated hardware and software to generate near-normal speech, complete with pitch, accent, and smooth transitions between sounds, mid to accept inputs from a variety of computers and specialized input devices. Commercial adaptations of Phase II prototypes would enable near-normal and even customized speech for a wide range of speech-impaired persons at costs comparable to those of today's devices. The sampled-speech subsystems developed in phase I and II could also be adapted and licensed for Integration Into computer, telecommunication and consumer products which require high-quality speech output.
