基于块稀疏信号的正则化自适应压缩感知算法

Regularized adaptive matching pursuit algorithm of compressive sensing based on block sparsity signal

在研究已有的块稀疏信号贪婪算法的基础上,提出一种正则化的自适应恢复算法。该算法在块稀疏度未知的前提下,添加了正则化的思想进行块挑选,从而更正确地挑选出块信号的支撑块,实现信号的重建。该算法首先在确定块的稀疏度和选择步长后,利用相关最大化原则实现支撑块的初次挑选;然后,依据已挑选出的支撑块再进行正则化分组,实现二次挑选;最终通过循环迭代正确挑选出整个信号的支撑块。通过仿真实验证明,该算法不仅不需要信号的块稀疏度作为先验知识,且较现有的块信号贪婪算法的重构概率更高,也比现有的块稀疏自适应贪婪算法所需的迭代次数更少和迭代时间更短。

A regularized adaptive matching pursuit algorithm as proposed after research and summarize the existing greedy algorisms based on block-sparse signal. This algorithm mainly in the light of regularized method under a condition that a block-sparse degree is unknown, so that the signal support set can be determined more accurately by the algorithm, then we can reconstruct a signal precisely. First, the algorithm initializes a sparsity degree and step size of a block signal; by maximizing the correlation between residual and measurement matrix, it realizes the selection of subset of the signal support. Then the algorithm updates the selected subset support set is acquired through iteration. The experimental results second prove th can get better reconstruction performance than other existing greedy algorith and it has less iteration number and iteration time than the other adaptive time. Finally, the exact at the proposed algorithm ms based on block signal, algorithm based on block signal.