一种新的贪婪回溯子空间追踪算法研究

发布时间：2018-02-20 23:12

本文关键词： 压缩感知贪婪算法重构回溯子空间追踪　出处：《计算机应用研究》2017年10期 　论文类型：期刊论文

【摘要】：针对目前的贪婪类算法在实际应用中出现的重构遮挡和虚假等问题,在分析该问题产生的原因基础上,提出了一种新的贪婪回溯子空间追踪(greedy backtracking subspace pursuit,GBSP)算法。该算法的基本思想是在每次的迭代过程中,采用回溯反馈和贪婪精选的思路进行支撑集选择。具体而言,在原子识别阶段,从残差投影中挑选出绝对值最大的K(K是信号稀疏度)个投影值位置,添加到候选支撑集中,为降低在此步骤中产生的错误概率,每次只将候选支撑集中的前s(sK)个最大值对应的位置添加到真实支撑集中进行更新;此后再进行投影计算和残差更新,直到完成支撑集的选择。新算法结合了正交匹配追踪算法和子空间追踪算法两者的优势,所以可较好地解决重构遮挡与虚假问题,使得压缩感知重构算法更具实用性。
[Abstract]:Aiming at the problems of reconstruction occlusion and falsehood in the practical application of the greedy class algorithm, the causes of the problem are analyzed. In this paper, a new greedy backtracking backtracking subspace pursuitto GBSP algorithm is proposed. The basic idea of the algorithm is to select the support set with the idea of backtracking feedback and greedy selection in every iteration process. In order to reduce the error probability in this step, pick out from the residual projection that the maximum absolute value KnK is the signal sparsity) and add it to the candidate support set, in order to reduce the error probability in this step, Each time, only the position corresponding to the first ssKs of the candidate support set is added to the real support set for updating; thereafter, projection calculations and residual updates are performed. The new algorithm combines the advantages of the orthogonal matching tracking algorithm and the subspace tracking algorithm, so it can solve the reconstruction occlusion and false problems, making the compression perception reconstruction algorithm more practical.
【作者单位】：湖北文理学院数学与计算机科学学院;
【基金】：湖北省教育厅科学技术研究项目(Q20142607)
【分类号】：TN911.7

【相似文献】