Abstract:We first discuss in the paper a two-step matching pursuit algorithm based on Ricker wavelet based on instantaneous characteristics of complex signal. Combined with high resolution time-frequency MPD, we propose a new method for thin bed thickness prediction, which is based on the one-to-one theoretical formulation between thin bed thickness and peak frequency in seismic signal even component on the theoretical model. The even and odd components in the centre of a window of seismic signal with zero phase wavelet are equivalent to the convolution of zero phase wavelet and the even and odd components of the corresponding reflection coefficients. The relationship between the peak frequency of odd component and bed thickness does not depend on reflection the coefficient of thin bed. So it is convenient to build a moldboard to predict the thickness of thin bed by substituting peak frequency calculated in a small window. Model test results prove the feasibility of the method.