Mariusz Ziółko (Poland)

Fourier-wavelet speech analysis

Speech representation in the time domain gives a little information about the speech signal properties. Hence, the appropriate transformation is an important problem to obtain the optimal spectral representation. Usually, methods which are based on the Fourier or wavelet transform are used. In this way, the frequency properties of speech are analysed. The Fourier transform gives averaged representation of a signal in the frequency domain. The wavelet transformation belongs to the group of frequency transformations and is used to obtain a time-frequency spectrum. Wavelet spectrum carries both, time and frequency representations. A hybrid system based on Fourier-wavelet transformations is suggested to obtain a new tool for speech analysis. For speech signal the wavelet spectrum is calculated first and next the Fourier transform is used to obtain Fourier-wavelet representations. The composition of these two transforms makes it possible to detect the specific voice signal properties.

Copyright © XXII PVC Organizing Committee 2013. All Rights Reserved.