Skip to Main Content
An automatic technique was developed to produce recognizable speech from a string of phonetic characters input from a teletype. The technique is based on modeling the acoustical consequences of the various configurations of the vocal tract. The simulation method has proven that speech generation from digital data is feasible and viable at a bit rate below 100 bits per second. The system takes a teletype input and, through a definitive set of algorithms executed on a PDP-11 computer, converts it to a form compatible with the input to a Digital Speech Synthesizer Model 4516, developed by the Rockland Systems Corporation. The Model 4516 is a time-multiplexed, cascaded, complex pole and zero network with a digital-to-analog (D/A) converter. The system was tested by listening to it and by the use of the Speech Analysis System, developed by the Aerospace Medical Research Laboratory. The Speech Analysis System consists of an analyzer and associated equipment necessary to produce a real-time, hard copy representation of the wave form characteristics of the speech input. The analyzer itself is called a COC filter and its design is based on the hydro-mechanical operation of the inner ear.