基于树型冗余字典正交匹配追踪的信号稀疏分解  

Sparse decomposition based on orthogonal matching pursuit with tree-structure redundant dictionary

作　　者：赵嵩[1] 马荣华[2] 薛朝改[3] 李恒建[4] 

机构地区：[1]郑州航空工业管理学院电子通信工程系,郑州450015 [2]郑州铁路职业技术学院软件学院,郑州450052 [3]郑州大学管理工程学院,郑州450001 [4]山东省计算中心计算机网络重点实验室,济南250014

出　　处：《扬州大学学报（自然科学版）》2011年第4期52-55,82,共5页Journal of Yangzhou University：Natural Science Edition

基　　金：国家自然科学基金资助项目(70971119);河南省教育厅自然科学研究计划项目(2011A510024)

摘　　要：结合树型结构和正交匹配追踪算法,提出一种信号稀疏分解的新方法.该方法的基本思想是在基于树型结构的匹配搜索过程中引入正交化过程,其中树型结构可以快速有效地实现稀疏分解,正交匹配追踪算法提高了信号分解的收敛速度和稀疏性.对语音和地震信号的测试实验结果表明,该算法能以较快的速度收敛到零.Tree based orthogonal matching pursuit is proposed to overcome the convergence of sparse decomposition.Sparse decomposition can be fast solved by tree based pursuit,however,the tree based pursuit is locally best in essence.This paper proposes the orthogonal matching pursuit algorithm that maintains full backward orthogonality of the residual（error） at every step and thereby leads to improved convergence.Also,it guarantees the sparsity of results and accuracy of reconstructed image.Speech signal and earthquake signal are tested via tree based orthogonal matching pursuit separately,both of which have better convergence performance than tree based matching pursuit.

关 键 词：稀疏分解 正交匹配追踪 冗余字典 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]