作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程 ›› 2014, Vol. 40 ›› Issue (12): 132-135. doi: 10.3969/j.issn.1000-3428.2014.12.024

• 人工智能及识别技术 • 上一篇    下一篇

基于前向预测策略的块正交匹配追踪算法研究

曾辉,胡凯,肖芬,徐真珍,黄紫橙   

  1. 湘潭大学智能计算与信息处理教育部重点实验室,湖南 湘潭 411105
  • 收稿日期:2013-12-12 修回日期:2014-02-22 出版日期:2014-12-15 发布日期:2015-01-16
  • 作者简介:曾 辉(1987-),女,硕士,主研方向:压缩感知,图像处理;胡 凯,讲师、博士;肖 芬(通讯作者),副教授、博士;徐真珍、黄紫橙,硕士。
  • 基金资助:
    国家自然科学基金青年基金资助项目(61202192)

Research of Group Orthogonal Matching Pursuit Algorithm Based on Forward Prediction Strategy

ZENG Hui,HU Kai,XIAO Fen,XU Zhenzhen,HUANG Zicheng   

  1. Key Laboratory of Intelligent Computing & Information Processing, Ministry of Education,Xiangtan University,Xiangtan 411105,China
  • Received:2013-12-12 Revised:2014-02-22 Online:2014-12-15 Published:2015-01-16

摘要: 在块正交匹配追踪算法中,候选集匹配原子块的选择策略对信号重建具有重要作用,但是该算法在迭代选择原子块的过程中,每次选择当次最优迭代的原子块,并不能保证最终迭代性能是最优的。为此,提出一个基于前向预测的最优匹配原子块选择策略的块正交匹配追踪算法,通过预测原子块在未来迭代过程中的性能选择最优匹配的原子块。实验结果表明,与块正交匹配追踪算法相比,该算法的重建误差随前向预测参数的增大而减小。

关键词: 压缩感知, 块稀疏信号, 前向预测, 块正交匹配追踪, 局部最优原子, 信号重建

Abstract: A set of candidate matching selection strategy with atomic blocks is crucial to the performance of the signal reconstruction for Group Orthogonal Matching Pursuit(GOMP) algorithm.However,in the process of iteration selection for atomic blocks,when select the local optimum atomic block in each iteration,it cannot guarantee that the final iteration is optimal performance.So this paper proposes the GOMP algorithm based on Look Ahead(LAGOMP).The selection of atomic block in the current iteration depends on its effect on the future iterations.Experimental results show that,the proposed algorithm performs much better than GOMP algorithm for the reconstruction error decreases with the increases to the forward prediction parameters.

Key words: Compressed Sensing(CS), block sparse signal, forward prediction, Group Orthogonal Matching Pursuit(GOMP), local optimum atom, signal reconstruction

中图分类号: