基于成比例归一化最小均方误差算法的压缩感知信号重构.pdfVIP

控制理论与应用 《自动化绞求与应用)) 2016 军第35 卷第 11 期 Contr划、曲ry and AppIìcations 基于成比例归一化最小均方误差算法的 压缩感知信号重构 吴莉 (阿坝师范学院数学与财经系，数学研究所，四川汶川 623∞2) 摘 要:在压缩感知理论框架中， 一个关键问题是信号的稀疏重构。这个问题可以归结为一个结构化的非光滑优化问题.本文 提出了一种基于自适应滤波成比例归一化最小均方的压缩传感重构算法ASS-ZA-LMS-SA(adaptive step 5ize zero- attraction 1臼5t m回n SQuare sigmoid 也orithm). 在该算法中，为了提高算法的收敛性，我们将创gmoid 函数引人到压 缩感知重构的代价函数中，此外，我们利用自适应滤波与信号稀疏重构的相似性，将成比例归一化最小均方算法应用到压 缩感知重构，设计了一种基于自适应滤波算法的压缩感知精疏信号重构算法.实验结果表明，与传统的压缩感知重构算法 相比，我们所提出的算法具有更高的重构精度以及更快的收敛速度. 关键词: 压缩感知$重构，自适应滤波$成比例归一化最小均方误差 中图分类号 :TP309.7 文献标识码 :A 文章编号:1 ∞3-7241(2016)11 -仪肌-06 Compressive Sensing Signal Reconstruction 8ased on Propo时ionate Normalized Least Mean Squares Algorithm WULi (Institute ofMathematics, Department ofMathematics,ABa Teachers College, Wenchuan 623002 China) Abstr act : 1n compressive sensing (CS) framework, sparse signal reconstruction is a key problem. This problem can be fonnulated as a structured non-smooth optimization problem . This paper proposes a CS reconstruction algorithm adaptive step size zero-a阳action least mean square sigmoid algorithm (ASS-ZA-LMS-SA) based on adaptive fìltering proportionate nonnalized least mean squares. In the proposed algoritbm , in order to improve tbe convergence performances , a sigmoid function is incorporated into the cost function of the CS reconstruction. Besides, based on the methodological similarity between CS reconstruction and adaptive fìltering , an adaptive fìltering CS reconstruction algorithm is derived by applying the proportionate nonnalized least mean square (PNLMS) into CS reconstruction . The exper