Third International Workshop on Models and Analysis of Vocal Emissions for Biomedical Applications (MAVEBA 2003)

Florence, Italy
December 10-12, 2003

An Efficient Method of Speech Signal Reconstruction Based on Neural Network and Fast Deconvolution Algorithm

A. M. Krot, H. B. Minervina, V. V. Sarapas

United Institute of Informatics Problems of the National Academy of Sciences of Belarus, Minsk, Belarus

In this paper we propose a new method of speech signal restoration based on a well-known fast deconvolution algorithm and a modern neural network approach. Such a combination inherits the adaptive capability from a neural network as well as the effective inverse filter calculation. According to our expectations, the experimental results reveal the fact that the new method is superior to the traditional ones.

Index Terms. Neural network, signal restoration, fast inverse deconvolution

Full Paper (reprinted with permission from Firenze University Press)

Bibliographic reference.  Krot, A. M. / Minervina, H. B. / Sarapas, V. V. (2003): "An efficient method of speech signal reconstruction based on neural network and fast deconvolution algorithm", In MAVEBA-2003, 73-75.