Edward E. David Jr.
Artificial speech -- speech generated by machine -- fascinated researchers as far back as the eighteenth century. People in that day were intrigued with the notion that human processes might be duplicated mechanically, and they knew that speech is uniquely human. Today we are interested in artificial speech not only out of scientific curiosity, but for many reasons--to create reading machines for the blind and voice answer-back by computer over telephone circuits, for example.
To make artificial speech, we must duplicate some of the functions of the human vocal tract, for it is within the vocal tract that the acoustic wave we hear as speech is generated. It begins with the lungs which are the primary source of energy for speech. In speaking normally, air is forced from the lungs and the resulting air stream carries this energy. This energy of flow is converted to sound by the vocal cords or by turbulence at a constriction formed in the vocal tract, for example at the teeth when making an "s" sound.
Actually, the sound produced by the vocal cords or by turbulence is easily duplicated mechanically. The vocal cord sound is merely a buzz, much like a "Bronx cheer" or other rude sounds. Many people who have lost their vocal cords through surgery use a transistorized artificial larynx to produce a buzzing sound for speaking.
This buzz does not sound much like speech--something more is needed. That something is provided by the remainder of the vocal tract formed by the tongue, palate, teeth, and lips. Together these form a tube through which the sound from the