Architecture Design of a High-Quality Speech Synthesizer Based on the Multipulse LPC Technique

This paper describes an architectural design of a high-quality speech synthesizer based on the multipulse linear predictive coding technique. The algorithms implemented in the design are derived. The main features of the synthesizer are a multiprocessor architecture, real time operation,
\mu-/A
-law coded PCM output, flexible system interface, relative ease of programming, high quality of synthesized speech, and variable bit rates. The synthesizer is highly testable as most of the internal resources can be addressed externally through a parallel system-controller port and a set of instructions.