ISCA Archive Eurospeech 1995
ISCA Archive Eurospeech 1995

Spectral envelope of speech using wavelets

Paul Micallef, Edward Chilton

The dyadic wavelet transform can be used to separate out the energy of a signal in various bands. The transform makes use of smoothing functions and wavelet functions that result in a very efficient method of signal decomposition and reconstruction. In this paper wavelet transform techniques are applied to separate out the spectral envelope from a Short Time Fourier Transform of a frame of speech. The results are compared to the spectral envelope estimation using LPC and cepstral methods.