稀疏信号重构的残差最小化追踪  

Sparse Signal Recovery via Residual Minimization Pursuit

作　　者：宋和平[1] 王国利[2] 

机构地区：[1]江苏大学计算机科学与通信工程学院,江苏镇江212013 [2]中山大学信息科学与技术学院,广东广州510006

出　　处：《信息与控制》2014年第6期722-726,734,共6页Information and Control

基　　金：国家自然科学基金资助项目(61375080;61202110;61170126);江苏省自然科学基金资助项目(BK2012284;BK20130528);江苏大学高级人才科研启动基金资助项目(12JDG050)

摘　　要：提出一种新的压缩传感稀疏信号重构算法——残差最小化追踪(residual minimization pursuit,RMP).残差最小化追踪RMP每次迭代选择残差信号在测量矩阵的正交投影绝对值最大的元素来检测支持集,然后求解支持集上的最小二乘解更新稀疏信号.另外,提出两种扩展残差最小化追踪RMP算法,算法每次迭代选择多个元素来检测支持集.实验结果表明,残差最小化追踪RMP稀疏重构性能优于正交匹配追踪OMP算法.A new sparse signal recovery ＂algorithm, dubbed as residual minimization pursuit （ RMP）, is proposed for compressive sensing signal reconstruction. This algorithm iteratively detects the support set of the true sig- nal by selecting the element with the laz＇gest magnitude of orthognnal projection of residual signal onto the measurements matrix, and then updates the unknown signal using a least-squares solution on the detecting support set. In addition, two support detection strategies are devised by spotting several elements in each iter- ation. The experimental studies a presented to demonstrate that the RMP algorithm offers an attractive alter- native to OMP for sparse signal recovery.

关 键 词：压缩传感 稀疏信号重构 贪婪算法 残差最小化追踪 

分 类 号：TN911.7[电子电信—通信与信息系统]