Abstract:Because of Heisenberg's uncertainty principle,the linear algorithms of time-frequency spectral decomposition,for example wavelet transform and generalized S transform,cannot simultaneously have high resolution in time and frequency domains.Overcoming the window function limit,the matching pursuit can precisely characterize signal features in time and frequency domains.This paper firstly discusses the algorithm theory of matching pursuit to seismic signal,and then put forward the reasonable improvement to standard Morlet wavelet,which can perfect the time-frequency atomic database,thus enhancing the precision of seismic signal matching decomposition.Based on the time-bandwidth of Morlet wavelet,the paper finally uses window treatment technology to dynamically scan matching wavelet atoms at the same time in multi-windows,so that calculation efficiency can be appropriately increasing.The model test and actual data analysis show that matching pursuit of seismic signal based on improved Morlet wavelet is more precise and efficient,and has some antinoise ability,suitable for quantitative analysis of the spectrum variation of seismic data,which will be helpful to study oil gas reservoir distribution.