Low‐cost speech synthesis is here and along with it the need for high compression of stored speech. A method for achieving speech compression to 1200 bits per second based on LPC parameters is presented. The low data requirement results from an efficient data coding procedure which incorporates a variable frame rate scheme. Which reflection coefficients are to be used and the parameter update intervals are determined using a dynamic programming procedure which maximizes frame length while minimizing a prediction residual energy ratio measure. Good quality speech can be obtained from this analysis procedure using the TMC0280 speech synthesizer. A demonstration of the speech quality will be presented.
This content is only available via PDF.
© 1978 Acoustical Society of America.
1978
Acoustical Society of America