基于回溯原子支撑集的广义正交匹配追踪算法研究  

Research on Generalized Orthogonal Matching Pursuit Algorithm Based on Backtracking Atom Support Set

作　　者：郝峥旭 杨斌 胡忞 易朋兴[1] HAO Zhengxu;YANG Bin;HU Min;YI Pengxing(School of Mechanical Science and Engineering,Huazhong University of Science and Technology)

机构地区：[1]华中科技大学机械科学与工程学院

出　　处：《仪表技术与传感器》2024年第5期74-78,84,共6页Instrument Technique and Sensor

基　　金：国家重点研发计划项目(2018YFB2003303)。

摘　　要：为解决广义正交匹配追踪算法(gOMP)在迭代过程中无法删除匹配错误原子问题,提出了一种新的算法:基于回溯原子支撑集的广义正交匹配追踪算法(BgOMP)。该算法在每次迭代选择原子前,首先检验支撑集中是否存在由先前迭代错误选择的原子,并据此更新原子支撑集。这一改进策略允许算法在每次迭代中回溯并纠正先前的错误。通过仿真实验,使用一维信号的成功重构概率和二维信号的峰值信噪比作为评价指标来验证BgOMP算法的性能,实验结果表明:与原始的gOMP算法相比,BgOMP算法显著提高了信号的重构性能。To address the issue of the generalized orthogonal matching pursuit algorithm(gOMP)being unable to eliminate erroneously matched atoms during its iterative process,a novel algorithm was proposed,that is the generalized orthogonal matching pursuit based on backtracking atom support set(BgOMP).Before selecting atoms in each iteration,the BgOMP algorithm initially examined whether the support set contained atoms incorrectly chosen in previous iterations and updated the atom support set accordingly.This enhanced strategy allowed the algorithm to backtrack and correct previous errors during each iteration.Simulation experiments were conducted to verify the performance of the BgOMP algorithm,utilizing the successful reconstruction probability of one-dimensional signals and the peak signal-to-noise ratio(PSNR)of two-dimensional signals as evaluation metrics.The results demonstrate that the BgOMP algorithm significantly improves signal reconstruction performance compared to the original gOMP algorithm.

关 键 词：压缩感知 重构算法 回溯思想 成功重构概率 峰值信噪比 

分 类 号：TN911[电子电信—通信与信息系统]