步长自适应ICA在地震信号去噪中的应用  

Application of adaptive step size independent component analysis to seismic signal denoising

作　　者：张银雪[1] 田学民[1] 

机构地区：[1]中国石油大学信息与控制工程学院,山东东营257061

出　　处：《计算机工程与应用》2011年第31期215-219,共5页Computer Engineering and Applications

基　　金：山东省自然科学基金(No.Y2007G49)~~

摘　　要：提出一种基于改进相对梯度独立分量分析(ICA)的叠前地震信号去噪方法。建立ICA的步长函数与信号相依性测度之间的非线性关系,根据信号的分离状态自适应调节步长函数;在分离矩阵的更新公式中引入动量项,削弱步长函数快速减小对分离矩阵更新的影响;将改进ICA模型应用于叠前地震信号去噪处理中。实验结果表明,改进的变步长ICA算法可以有效地压制叠前地震记录中的随机噪声,并且能很好地保护有效地震信号。An improved approach based on the modified relative gradient Independent Component Analysis（ICA） is proposed for prestack seismic denoising in this paper.Firstly,this method establishes a nonlinear relationship between the step size function and the measure of signal dependence.This algorithm can adjust the step size function adaptively.On the other hand,it adds a momentum to the separating matrix.The momentum can diminish the effect of the step size function＇s sharp reduction.Finally,it uses the modified ICA model to reduce prestack seismic random noise.Experimental results demonstrate that the proposed ICA method has good effect on removing random noise in prestack seismic record.In addition,the proposed method can protect the available seismic signal effectively.

关 键 词：叠前地震记录 去噪 独立分量分析 相对梯度算法 变步长 

分 类 号：TN911.7[电子电信—通信与信息系统]