基于独立分量分析的多次波盲分离方法研究  

A method of multiple blind separation based on ICA

作　　者：郭科[1,2] 郭思[3] 何国柱[4] 陈辉[1,2] 罗德江[1,2] 刘诚[1] 杨威[1] 

机构地区：[1]成都理工大学信息管理学院,成都610059 [2]数学地质四川省高校重点实验室,成都610059 [3]成都理工大学沉积地质研究院,成都610059 [4]四川农业大学,都江堰611830

出　　处：《地球物理学进展》2010年第3期1075-1080,共6页Progress in Geophysics

基　　金：油气藏地质及开发工程国家重点实验室开放基金项目(PLC200703);国家自然科学基金项目(40873035);教育部博士点基金项目(200806160015);四川省教育厅自然科学项目重点项目(2006A114)联合资助

摘　　要：独立分量分析是一种以高阶统计量为基础的信号处理方法,它以分离非高斯混合信号为目标,建立起各个分量的统计独立性判据.在传统方法中,多次波压制技术是基于二阶统计量,求得最优解得基本前提是一次波和多次波正交.本文将ICA应用于多次波问题,在对地震数据基本构成作了分析的基础上,建立起多次波盲分离的ICA模型,并对其假设条件和固有不确定问题进行了详细的研究分析,然后给出了基于负熵的快速ICA算法并加以改进,最后进行仿真实验.试验结果表明,该方法能有效地压制地震资料中的多次波信息,较好的恢复一次波信息.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.

关 键 词：独立分量分析 多次波压制 高阶统计量 快速ICA 负熵 

分 类 号：P631[天文地球—地质矿产勘探]