The Application Of Multiple Attenuation Based On Independent Component Analysis

Recently, Independent component analysis (ICA) is getting more and more widely used in digital image processing, pattern recognition, data compression, speech signal processing and economic field. As a new multidimensional signal processing method, it derived from blind source separation (BSS). ICA depends on the statistical properties of the input signals, it could recover the unobserved source signals without any prior information given only the senson observations that are unknown linear mixtures of the independent sources.At present, the traditional technology for multiple attenuation based on the principle of the least energy of output signal are deduced from second order statistical data. To secure the best solution, a prerequisite is necessary, that is the primary and multiple of seismic signals must be orthogonal. But it's not in accordance with the real situation of seism exploration. And ICA which is based on high-level statistics could solve this problem. But we need to assume that there's no differences in waves between the forecast and actual multiple before using ICA in the adaptive subtraction. In this paper, we eliminated the differences between the forecast and actual multiple by preprocessing the forecast multiple with the equipoise pseudomultichannel matching filter. Then consider to the statistical characteristics of the seism data, a signal model which could be processed by using ICA to multiple attenuation is set up. And settles the two ambiguities of the amplitude and the permutation of the independent components by adopting some prior information in the mixture matrix ingeniously. Then design algorithm to carrie out the separation of the primary and the multiple successfully by use of the non-gaussian algorithm.From the results of processing both simple model and real seismic data by the algorithm of this paper, we can see that the primary could be almost completly recovered.