Speech recognition for the disabled

A simple and inexpensive system for telephone communication that includes a modular speech recognition device developed for use by a disabled person is presented. User considerations are reviewed, and hardware details are given. The complete automatic speech recognition (ASR) system consists of five main blocks: a TCM29C18 codec (analog interface for the digital signal processor), an SRS-1 speech recognizer, an LED display, a dual-tone multifrequency generator, and tone enable plus data input circuitry. Results of a semiqualitative assessment in terms of recognition success for the limited vocabulary of numbers and commands from the user are reported. The results are described in chronological order, beginning with testing of the SRS-1 and the codec working as a stand-alone recognizer system, and leading to the testing of the recognition system combined with the tone generator. Finally, results were obtained when the whole ASR system was interfaced to the telephone line. Some further work is suggested.<>