A Feasibility Study of Very Low Rate Speech Compression Systems
Abstract
This report documents the results of a feasibility study for very low rate speech compression systems. Two systems were implemented: (1) A variable rate, hybrid formant-LPC vocoder transmitting at an average of 500 bps, and (2) a phonetic vocoder transmitting at an average rate of 75 bps. The intelligibility of the 500 bps system was judged to be high, but with a loss in naturalness and speaker identifiability. The 75 bps system utilized a phonetic recognition program and transmitted phonemes, pitch and duration. The synthesis at the receiver was performed using a synthesis-by-rule program, with a resulting good intelligibility but machine-like quality. In a listener judgment experiment of the vocoder, a recognition accuracy of over 80% was deemed necessary for communication purposes. A study was then undertaken to determine the feasibility of a real-time 75 bps vocoder. It was concluded that such a system was indeed feasible, with over 80% phonetic recognition, natural sounding synthesis, and real-time operation using off-the-shelf technology.
Document Details
- Document Type
- Technical Report
- Publication Date
- Feb 01, 1977
- Accession Number
- ADA044400
Entities
People
- C. Cook
- D. Klatt
- J. Makhoul
- Robert E. Schwartz
Organizations
- BBN Technologies