
Sixth International Conference on Spoken Language Processing
(ICSLP 2000)
Beijing, China
October 1620, 2000 

ReEstimation of LPC Coefficients in the Sense of L&Inf; Criterion
Zhang Sen, Katsuhiko Shirai
Depart. of Information and Computer Science,
School of Science and Engineering,
Waseda University, Tokyo, Japan
Now the generally used approaches such as
autocorrelation method and covariance
method for estimating LPC coefficients are to
solve a set of linear equations by using of
LevinsonDurbin recursion or lattice
formulations, but the LPC coefficients
computed are best only in the sense of L2
criterion. For speech processing, L&inf;
criterion is a more suitable measure metric.
The idea of our approach is: from the initial
values of LPC coefficients, the residual errors
could be reduced step by step by using Least
Squares process iteratively until the LPC
coefficients are approximately best in the
sense of L&inf; criterion. Furthermore, this
approach could be applied to other problems
for estimating some parameters.
Full Paper
Bibliographic reference.
Sen, Zhang / Shirai, Katsuhiko (2000):
"Reestimation of LPC coefficients in the sense of l&inf; criterion",
In ICSLP2000, vol.1, 737740.