Second International Workshop on Models and Analysis of Vocal Emissions for Biomedical Applications (MAVEBA 2001)

Florence, Italy
September 13-15, 2001

Fast Algorithms for Wiener Kernels Computing in Speech Phoneme Recognition

Alexander M. Krot (1), Polina P. Tkachova (2)

(1) Institute of Engineering Cybernetics of the National Academy of Sciences of Belarus, Minsk, Belarus
(2) Belarusian State University, Minsk, Belarus

This paper presents the nonlinear speech phoneme decomposition based on Volterra-Wiener functional series. It is shown the usage this nonlinear decomposition in speech recognition systems constructing. The fast algorithms for finding estimation of Wiener kernels in frequency domain permit to reduce essentially computing expenses for evaluation of signals decomposition.

Index Terms. speech recognition, nonlinear signal decomposition, Volterra-Wiener functional series, Wiener kernels measuring, fast algorithms

Full Paper

Bibliographic reference.  Krot, Alexander M. / Tkachova, Polina P. (2001): "Fast algorithms for Wiener kernels computing in speech phoneme recognition", In MAVEBA-2001, 246-253.