Speech Compression and Synthesis.
Quarterly progress rept. no. 5, 8 Jun-7 Sep 79,
BOLT BERANEK AND NEWMAN INC CAMBRIDGE MA
Pagination or Media Count:
This document reports on work towards a very low rate phonetic vocoder, text to speech, and multirate speech compression. This work included improvement of the phonetic synthesis algorithms and continued gathering of the diphone templates data base for phonetic synthesis. It also included the initial design of a phonetic recognizer to operate in conjunction with the synthesizer. The combination of these two programs will result in the very low rate vocoder. The method of spectral harmonic deviation was tested in an LPC vocoder environment. The sources for the MIT text-to-speech system were translated from BCL to BCPL, and the runable program was also transferred to our computer. Research was begun on a multirate speech compression system capable of operating over a wide range of data rates. Keywords include Speech synthesis phonetic synthesis diphone LPC synthesis vocoder speech compression linear prediction voice-excited coder high-frequency regeneration spectral duplication phonetic vocoder spectral template speech recognition and phoneme recognition.