SPEECH SYNTHESIS.

This paper describes a device that is capable of synthesizing speech in real time and is based on the digital signal processing chip. The device performs a function of a twelfth-order linear prediction coding synthesizer, and as such represents a linear dynamic system approximation of the vocal tract. In this model, short time segments of the speech waveform are derived as output of a system driven by a pseudo-periodic impulse sequence for voiced sounds, or by white noise for unvoiced sounds. The time-varying nature of the system is derived from the input information presented to the device for every new pitch from the input information presented to the device for every new pitch period. Interfacing of the device to standard microprocessors is easy, so that the synthesizer can conveniently be integrated into large systems.