The Techniques Of Adaptive Filter In Frequency Domain And Its Application

With the progress of physics, mathematics and others, especially computerscience, the technique of seismic prospecting is rapidly develop, and it is aimportant role in the field of oil rand gas prospecting and gain the bestachievement. There are there steps in the seismic prospecting: collection, dataprocessing and interpretation. Data processing in these steps is very important.Noise elimination is the basement of the seismic data processing. If it not workwell, the latter things should be affected. The high signal-to-noise rate (SNR)is the aim that people are pursuing for gaining good seismic profile and geologicinterpretation.There are all kinds of coherent noise in the seismic data, because the theearth's surface, particularity of construction and the condition of geology, itreduce the SNR of the seismic data. in the noises, the refraction wave is one ofthe coherent noises, and it can effect the distinguish of effective wave. If it is notbe cleared up in the seismic data, it can effect the correct interatation in theseismic profilThe aim of this technique is heighten the signal-to-noise rate (SNR) inseismic profile. The least mean square (LMS) filter in frequency domain canremove refraction effectively in the seismic data.B.Widrow and Hoff put forward the arithmetic of the least square (LMS).LMS arithmetic was applied widely, because it is easy to realize and isrobustness to the changes of statistic characteristics. LMS arithmetic was basedon the winer filter of minimum mean square error criterion (MMSE) andsteepest descent method. It is easy to calculate and reach the satisfactoryperformance. But its flaws are slowness of velocity of astringency andsensitiveness of the divergence of correlation matrix eigenvalue.Adaptive LMS filter in frequency domain is that the signal can be changedto the field of frequency firstly, and adopt some transforms to adjust the filtercoefficient. The paper discussed the foundation of mathematics of adaptive LMSfiler and applied it on the seismic data processing. There are three merits:1.According the filter result to adjust the filter coefficient, it makes filterperformance good and protect the effective signal.2.The count of process the data decreases greatly comparing with count intime domain, because convolution was replaced by commutation multiply. Tothe simplicial gradient descent, the astringency of adaptive process wasimproved.3.The processing in frequency domain can easily divide into differentblocks. The blocked adaptive LMS filter made blocked LMS arithmetic in thetime domain through fast algorithm in the frequency domain, and the filterweight coefficient of the same block hold the line and estimate the gradientdecent. When amending, it can adjust weight coefficient with higher precision.This paper apply the adaptive LMS filter to process the seismic data,according the model and practice data, we can see that this method has goodeffect and value of application, at the same time, the normal wave field was notdestroyed. It can heighten the SNR, enhance the continuity of reflected signaland improve the quality of the seismic profile.From this study, we can know that there are many methods of high SNR andresolution, the methods in this paper can obtain good results. At the same time,we should study the choosing of parameter more to obtain better methods thatcan get good results.