Research On Spectral Decomposition Of Seismic Signal Processing

Currently spectral decomposition technique of seismic signal is a common and effectivemethod in geophysical exploration. There are many methods of spectral decomposition ofseismic signals, such as the WVD, short-time Fourier transform and wavelet transform. Eachmethod has its own characteristics and limitations. In this paper, matching pursuits which is aflexible self-adaptive algorithm is used in spectral decomposition of seismic signals.Matching pursuit algorithm has flexible self-adaptation.The algorithm in this paperdecomposed the signal into a linear sum of time-frequency atoms which best match the signal.And then calculate the WVD of each atom and linear superposition. Pile up thetime-frequency distribution of all the atoms to constitute the time-frequency distribution ofentire signal. This method can effectively avoid the cross items of WVD and get a goodtime-frequency distribution.Due to the large amount of calculation of matching pursuit algorithm, hindered its wideapplication of signal processing. In order to solve this problem, traditional matching pursuitalgorithm has been improved in this paper. Using genetic algorithm to realize each timematching pursuit algorithm in the process of finding the best atomic, to reduce thecomputational complexity and improve the operation rate. Due to the genetic algorithm hasthe problem of slow convergent speed and Local optimum. In this paper, the geneticalgorithm has been improved. This paper introduced an improved adaptive genetic algorithmby adaptively changing the crossover probability and mutation probability to avoid thephenomenon of local optimum in traditional genetic algorithm. Eventually, the improvedmatching Pursuit algorithm will have a better convergence. The prior condition of seismicsignal, this paper introducing FFT, by changing the atomic library index range, change theinner product range, to reduce the amount of calculation, to further improve the operationspeed of the matching pursuit algorithm. At the end of the paper combines multiple seismictrace together, constitute a three-dimensional data volume with seismic trace, time andfrequency. Cutting the three-dimensional data volume to the frequency slices can obtain thetime-frequency distribution of the strata at different frequency. The presence of hydrocarboncan be instructed by the “low-frequency shadow” phenomenon of spectral energy at differentfrequencies.