Development of a programmable time-domain speech-synthesis system

Abstract

SPEECH-SYNTHESIS - the artificial generation of that series of sounds known as 'speech' - has advanced beyond the novelty stage to become a real alternative to the simple audible and visible indicators and displays common to so much of modern society. There is very little that can be done with a light or buzzer that can't be done better with a 'spoken' word. 'Toys' such as Texas Instruments Speak and Spell have been recognized as effective learning tools. Through the electronic mouths of those machines, children are exposed to new words in an exciting and interactive way. Talking calculators (Williams, 1994: 67) and timepieces have expanded the horizons of the blind. Pilots and drivers are relieved of the need to watch their meters and gauges continuously as alarms can be given with instructions as to what actions should be taken. Speechsynthesis devices are finding a myriad of uses in communications, appliances, automotive applications, clocks, instrumentation, language translators etc. (Savon, 1982: 62). Despite the obvious advantages speech-synthesis provide, many systems are limited in their application possibilities due to one or more of the following disadvantages: • expensive • inflexible • difficult to program • limited to an Anglo-Saxon language • resultant speech is unnatural and 'mechanical' in nature • speech synthesizer is difficult to control and require a microprocessor . a thorough phonetic knowledge is a prerequisite for programming the system