First International Conference on Spoken Language Processing (ICSLP 90)
This paper describes the recognition and synthesis strategies employed in the design and development of a voice input-output system using isolated words. The criterion for the design of the system have been low cost and its suitability for applications in the areas such as control of elctronic/mechanical systems and aids for handicapped by voice commands particularly using isolated words (Abdul Mobin et. al  ). The system consists of a speech recognition and a synthesis unit which can be used either separately or in combination. The hardware and software of the system are based on a Z-.80A microprocessor. The. recognition unit accepts isolated words spoken by a person through a close talking microphone. The detected signal is passed through a specially designed signal conditioning circuitry to obtain a constant signal level. The speech waveform is then processed to obtain a shortterm Fourier spectrum in real time using a audio spectrum analyzer chip ASA-16. The signal obtained from the analyzer is sampled and digitized at the. rate of 200 frames per second. The data is checked for background noise, impulse sounds, amplitude levels, minimum and maximum word lengths etc. Special software has been developed for amplitude normalisation, word boundary detection, time normalisation, data compression and pattern matching etc. ,  and . The speech synthesis unit has been designed using a commercially available phoneme based speech synthesizer chip SC-02 which is also controlled by the same Z-80A microprocessor. Programmes have been developed to synthesize Hindi and English words by generating a proper sequence of phonemes and controlling parameters such as speech rate, pitch, pitch movement rate, amplitude, articulation rate, vocal tract filter response, phoneme duration etc.
Bibliographic reference. Mobin, Abdul / Agrawal, S. S. / Kumar, Anil / Pavate, K. D. (1990): "A voice input-output system using isolated words", In ICSLP-1990, 1381-1384.