摘要
研究模型压缩感知中的块稀疏信号重构问题.在l2/l1模型基础上,提出一种基于交替方向法的块稀疏信号重构算法.在该算法中,首先对目标函数进行变量分裂,然后利用交替方向法对各变量进行交替更新,直至满足收敛条件.仿真实验中,将该算法与块正交匹配追踪和块压缩采样匹配追踪算法进行比较,结果表明该算法能够在保持高重构精度的前提下获得更快的计算速度.
The problem of block sparse signal reconstruction in model-based compressing sensing was studied.Applying alternating direction method,an efficient block-sparse signal recovery algorithm based on l2/l1 reconstruction model was proposed.In this novel algorithm,the objective function was transformed through variable splitting and four variables were alternately updated in the framework of ahernating direction method until the prespecified convergence criterion was satisfied.In computer simulation test,two state-of-art fast algorithms,named as block orthogonal matching pursuit and block compressive sampling matching pursuit,as well as the proposed algorithm were comprehensively compared and from the experimental results we could find that our algorithm was superior to the other two algorithms with respect to computational efficiency and estimation accuracy.
出处
《安徽大学学报(自然科学版)》
CAS
北大核心
2014年第2期61-67,共7页
Journal of Anhui University(Natural Science Edition)
基金
国家自然科学基金资助项目(61272333)
安徽省自然科学基金资助项目(1208085MF94)
国防预研基金资助项目(41101040402)
关键词
块稀疏信号重构
交替方向法
块坐标下降法
算法分析
block-sparse signal recovery
alternating direction method
block coordinate descent method
algorithm analysis