Seismic Signal Denoising Research Based On Generalized S Transformation And Threshold Function

Seismic exploration is an important means for identifying the undergroundgeological structure and surveying oil,natural gas resources and solid mineralresources,through observation and analysis of the earth on the artificial seismic waveresponse. In seismic prospecting, random noise always come along with effectivesignal in seismic records. It reduces the signal to noise ratio of seismic data, thedynamic correction velocity analysis, and the seismic imaging. Seismic exploration isa process,which is associated with the noise repeated struggle. Extracting usefulsignals from noise background is a seismic signal processing difficulty and heat, andsolving the problem will produce tremendous active effect for the practicalproduction.Seismic random noises are nonlinear, non Gaussian, nonstationary. Signalfrequency is changing over time. The classical Fourier transform, can only describethe signal from the frequency domain generally,this is suitable for the analysis of thedeterministic signal whose frequency components do not change with time. Seismicrandom noise frequency components is always changing over time, the traditionalFFT can only display frequency composition of the signal, but can’t show the time ofthem, so it can not effectively detect frequency changes with time of nonstationarysignal.It is difficult to analyze the signal local character. So traditional FFT method isnot applicable for the frequency analysis of seismic random noise. Time frequencyanalysis is an effective seismic random noise processing method, and a powerful toolfor the analysis of nonstationary signals. It provides time and frequency of jointdistribution information, and describes frequency of the signal changing with time.S transform is a modern time-frequency analysis tool developed in recentyears,and it absorbs the characteristics of Fourier transform and continuous wavelettransform and developed them.It uses the local Gauss functions whose scale canchange, and the time-frequency resolution can change with the frequency.Thepositive and the negative transformation can be achieved through FFT. We can use Stransform to analysis time-frequency characteristic of seismic wave and maketime-frequency adjustment artificially.Generalized S transform absorbed theadvantages of the short-time Fourier transform and wavelet transform. Compared withthe original S transform, new parameters λ and p are introduced. Appropriateadjustments are made for S transform window function,according to the frequencychanges. The problem that the time-frequency window cannot change is solved. And compared with wavelet transform, it does not need to satisfy the admissibilitycondition of wavelet. The time-frequency window can automatically adjust accordingto frequency, which is more adapted to the complex signal analysis. However, Stransform denoising process also has some defects.When time-frequency domain filteris being structured, some effective signal lose during the denoising process.In view of the problem above, in this paper, we introduce a compromisethreshold function to adjust the parameters of the S transform,and we combine thetwo ways to suppress seismic random noise, in order to obtain the best treatmenteffect, and less loss of information, to maximize the signal to noise ratio. According tothe above method, a large number of simulation experiments had been done, includingsingle-axis signal, multiple-axis signal denoising experiment, as well as the actualseismic data denoising experiment. The S transform denoising results and generalizedS transform with threshold function de-noising results, are compared. The resultsshow that, S transform and threshold function de-noising has better denoising effect.