Abstract:

Short-time Fourier transform is founded on the base of stationary signal analysis and can provide information in frequency domain with a bad ability of signal analysis in time domain. S transform is one of methods of signal analysis in time frequency domain, which is better than short-time Fourier transform, and can provide information of signal in time frequency domain. By using S transform to analyze Magnetotelluric（MT） sounding data in time frequency domain, it is helpful to filter noise in MT data and to improve the quality of MT data spectrum. This paper studied the method of time frequency domain filtering for suppressing noise in magnetotelluric sounding data by the generalized S transformation. Time frequency characteristics of noises of all kinds of wave and theirs influence are analyzed. According to the MT data processing characteristics and GST spectrum, time frequency filter and filter method are studied, which is suitable for suppressing the MT noises. Through actual MT data processing, the quality of impedance tensor estimate is improved and validity of the proposed method is verified.

Key words: S transform, magnetotelluric sounding, timefrequency window, noise