中国邮电高校学报(英文) ›› 2012, Vol. 19 ›› Issue (6): 123-128.doi: 10.1016/S1005-8885(11)60327-0

• Others • 上一篇    

Recovery of correlated row sparse signals using smoothed L0-norm algorithm

刘雨1,马骢1,朱旭琪1,张琳2   

  1. 1. 北京邮电大学
    2.
  • 收稿日期:2012-05-14 修回日期:2012-11-06 出版日期:2012-12-31 发布日期:2012-12-14
  • 通讯作者: 刘雨 E-mail:liuyurainy@gmail.com
  • 基金资助:

    This work was supported by the National Natural Science Foundation of China (61201149), the 111 Project (B08004), and the Fundamental Research Funds for the Central Universities.

Recovery of Correlated Row Sparse Signals Using Smoothed L0-Norm Algorithm

  • Received:2012-05-14 Revised:2012-11-06 Online:2012-12-31 Published:2012-12-14
  • Contact: Yu LIU E-mail:liuyurainy@gmail.com
  • Supported by:

    This work was supported by the National Natural Science Foundation of China (61201149), the 111 Project (B08004), and the Fundamental Research Funds for the Central Universities.

摘要:

Distributed compressed sensing (DCS) is an emerging research field which exploits both intra-signal and inter-signal correlations. This paper focuses on the recovery of the sparse signals which can be modeled as joint sparsity model (JSM) 2 with different nonzero coefficients in the same location set. Smoothed L0 norm algorithm is utilized to convert a non-convex and intractable mixed L2,0 norm optimization problem into a solvable one. Compared with a series of single-measurement-vector problems, the proposed approach can obtain a better reconstruction performance by exploiting the inter-signal correlations. Simulation results show that our algorithm outperforms L1,1 norm optimization for both noiseless and noisy cases and is more robust against thermal noise compared with L1,2 recovery. Besides, with the help of the core concept of modified compressed sensing (CS) that utilizes partial known support as side information, we also extend this algorithm to decode correlated row sparse signals generated following JSM 1.

关键词:

DCS, JSM, row sparse signal, smoothed L0-norm, partially known support

Abstract:

Distributed compressed sensing (DCS) is an emerging research field which exploits both intra-signal and inter-signal correlations. This paper focuses on the recovery of the sparse signals which can be modeled as joint sparsity model (JSM) 2 with different nonzero coefficients in the same location set. Smoothed L0 norm algorithm is utilized to convert a non-convex and intractable mixed L2,0 norm optimization problem into a solvable one. Compared with a series of single-measurement-vector problems, the proposed approach can obtain a better reconstruction performance by exploiting the inter-signal correlations. Simulation results show that our algorithm outperforms L1,1 norm optimization for both noiseless and noisy cases and is more robust against thermal noise compared with L1,2 recovery. Besides, with the help of the core concept of modified compressed sensing (CS) that utilizes partial known support as side information, we also extend this algorithm to decode correlated row sparse signals generated following JSM 1.

Key words:

DCS, JSM, row sparse signal, smoothed L0-norm, partially known support

中图分类号: