Voice or speech synthesizers combine hardware and software into systems that let your computer talk to you. Computers and Electronics and Radio Electronics (Mar.83) covered two methods of producing human voice sounds on ZX/TS computers. They are: Pulse Code Modulation and Linear Predictive Coding.
Pulse Code Modulation (PCM) uses silicon chip recordings (digitized) of a voice speaking selected words. A human voice encodes words in a ROM chip resulting in lifelike sounds. PCM’s disadvantage is you must have ROM addresses for all desired words.
Linear Predictive Coding (LPC) feeds a combination of random and periodic signals to the electronic voice tract. These periodic and random signals are modulated digitally to produce unique individual sounds. But, these sounds have an unnatural, mechanical ring to them.
Voicetech sells units that use the LPC technique (covered Mar. 83, Radio Electronics). R.I.S.T.‘s Parrot includes Voicetech’s evaluation kit and necessary ZX/TS hardware. These units produce sound by concatenating allophones (individual unique sounds) into an intelligible word or phrase.
Which technique of speech synthesis is better? Pulse Code Modulation produces a better sound, but costs more both in dollars and memory space. Current versions of Linear Predictive Coding are hard on untrained ears (it’s similar to listening to someone with a heavy accent). But, LPC has potential for infinite speech and the quality will no doubt improve.
As the micro market grows more competive and peripheral prices drop coupled with increased memory size of most personal computers, both will become viable methods of speech synthesis.