A method of multiple blind separation based on ICA

ICA is a new multi-dimensional signal processing method based on high-order statistics,which is used to separate non-Gaussian mixed signal,and set up a criterion to decide whether all the components are statistically independent.In the traditional method,multiple suppression technique is based on second-order statistics,which requires that signal primary reflection and multiple must be orthogonal for optimality solution.In this paper,ICA was applied to the multiple suppression issue.We sets up the ICA model of multiple blind separation after analyzing the basic composition of seismic data,and give a detailed analysis of the assumptions and inherent uncertainties of the issue.Then we present the fast ICA algorithm based on negentropy and improve it.Experimental results show that this method can attenuate multiple effectively and recover signal reflection preferably.