Skip to Main Content
Due to its programmability, the Programmable Digital Signal Processor (PDSP)chip set lends itself to the implementation of a variety of speech synthesis functions. When programmed as an all-pole lattice filter used in LPC synthesis, the PDSP allows the user to exploit quality-enhancement techniques such as mixed sourcing, pitch- or time-synchronous updating, variable frame and filter length, interpolation, various quantizing schemes, and adjustable speaking rate. Since the quantizing algorithm is under user control, the encoded bit rate may range from 1100 to 28,000 bits per second according to speech quality requirements. The paper discusses these techniques and a commercially available LPC speech synthesis circuit board which illustrates typical PDSP application.