Speech Compression and Synthesis
Abstract
This document reports on work toward a very low rate phonetic vocoder and a multirate speech compression system. The phonetic vocoder consists of a phonetic recognizer based on a trained diphone network, and a natural phonetic synthesizer which also uses diphone templates as a model for speech. This quarter, the basic data base of transcribed diphones for the phonetic synthesizer was completed. There are currently 2652 diphones. The initial versions of the diphone network compiler and network matcher programs for phonetic recognition were also implemented. On the multirate speech compression system, further modifications and improvements to the basic baseband ATC algorithm were tried. Keywords include: Speech synthesis; Phonetic synthesis; Diphone; LPC synthesis, vocoder; High-frequency regeneration; Spectral duplication; and Phoneme recognition.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 1979
- Accession Number
- ADA155467
Entities
People
- J. Klovstad
- J. Makhoul
- J. Sorensen
- M. Berouti
- Robert E. Schwartz
Organizations
- BBN Technologies