Non Linear Approaches to the Detection of Discontinuities in Concatenative
Speech Synthesis
Yannis Pantazis, Yannis Stylianou
Department of Computer Science, University of Crete
P.O.Box 2208, Heraklion, Crete, GR-714 09 GREECE
Phone:.+ 30 2810 393 526 .. Fax: + 30 2810 393 501
E-mail:. {pantazis,yannis}@csd.uoc.gr
An objective distance measure which is able to predict audible discontinuities in
concatenative speech synthesis systems is very important. Previous results showed
that linear approaches are not very effective to detect audible discontinuities—the
best result was the Kullback-Leibler distance on power spectra with rate 37.1%. In
this paper, we present two nonlinear approaches for the detection of discontinuities.
The first method is based on a nonlinear harmonic model for speech while the second
method is based on the demodulation of speech in an amplitude and a frequency component
using the Teager energy operator. Preliminary results show that detection rate can
reach 60% or little more, which is an improvement of about 75% over previous published
results. In an attempt to improve the detection rate we are now working in the
combination of the above methods.