In order to evaluate eathquake site amplification characteristics, horizontal to vertical Fourier spectrum of microtremor has been widely used. As long as the Fourier transform (FT) cannot distinguish between stationary and non-stationary coefficients we cannot eliminate the contamination of microtremors signals from short period transients. The wavelet transform (WT), using the property of localization of wavelet bases has been widely used in signal processing. Unlike the Short Time Fourier Transform (STFT) in which the width of window is fixed, the WT localizes signal in a variable window using the dilation parameter. This property, which derived directly from multiresolution analysis provide us the ability to decompose a signal in a well localized set of coefficients and identify the non-stationary portions of it. In the present study we use the WT in order to eliminate the non-stationahties in microtremor signals before we calculate the spectrum of each one using conventional FFT algorithms
(EN)
