Abstract:

The introduction of the wavelet transform (WT) and the analysis of multiresolution around the end of 1980, assumed an increase in the interest in signal processing and analysis. The development of the WT can be thought of as an attempt to improve the traditional analysis accomplished through STFT. However, STFT and WT outline resolution problems too. Attempts at improving the WT are the wavelet packets, and the designated M-band wavelet analysis. These analysis methods are similar in that they do not consider the characteristics of the signal for purposes of analyzation. A method that takes into account these characteristics is the ``matching pursuits'' of Mallat and Zhang. This method selects from the dictionary of time-frequency atoms those atoms that are more representative in the signal. Introduced in this work is an analysis method that is adapted to spectral characteristics of the signals. First, the spectral characteristics are determined. The transform of adjustment of the window size (TAWS), is an ``interpolation'' between the WT analysis and the STFT. This interpolation is accomplished by taking into account the signal characteristics: If frequencies that are very well defined do not exist in the spectrum, the analysis is the WT.