Title :
Simplified fast least-squares algorithms for speech predictors
Author :
Evci, C.C. ; Bellanger, M.G.
Author_Institution :
TRT, Paris, France
fDate :
5/12/1988 12:00:00 AM
Abstract :
The letter deals with the development of a simplified fast least-squares algorithm which is free of roundoff error accumulation. The simplified algorithm requires 9N MADPR (multiplications and divisions per recursion) rather than 10N MADPR as in the fast least squares or fast Kalman (FLS) case where N is the order of predictor. The superiority of SFLS over FLS and LMS approaches is illustrated by prediction gain performance curves for various speech signals
Keywords :
filtering and prediction theory; least squares approximations; speech analysis and processing; MADPR; fast least-squares algorithms; prediction gain performance curves; speech predictors; speech signals;
Journal_Title :
Electronics Letters
