基于MS-ROMP信号重构算法的欠定盲分离

为了提高最小支撑正交匹配追踪(least support denosing-orthogonal matching pursuit, LSD-OMP)算法的重构精度,缩短重构时间,改善算法性能,提出一种基于多重支撑的正则化正交匹配追踪(multiple support of regularization orthogonal matching pursuit, MS-ROMP)算法。由于LSD-OMP算法仅选择一些原子来定位支撑集,并且无法消除添加到支撑集中的错误原子,因此信号恢复精度降低并且重构时间增加。针对此问题,本文通过改进算法终止条件,引入多重支撑和正则化来改善算法性能,即通过设置阈值,剔除一些错误的原子,并组合一些支持集来定位最佳支持集,从混合信号中分离出源信号,从而更加精确的实现欠定盲源分离。仿真实验验证了该算法的有效性。

Regularization orthogonal matching pursuit based on multiple support

In order to improve the reconstruction accuracy of the least support denosing orthogonal matching pursuit (LSD-OMP) algorithm, short the reconstruction time and improve the performance of the algorithm, a regularized orthogonal matching pursuit based on multiple support (MS-ROMP) is proposed. Since the LSD-OMP algorithm selects only a few atoms to locate the support set and cannot eliminate the wrong atoms added to the support set, the signal recovery accuracy is reduced and the time is increased. To solve this problem, the performance of the algorithm is improved by improving the termination condition and introducing multiple support and regularization. By setting the threshold, eliminating some wrong atoms, and combining some support sets to locate the optimal support set, the source signal is separated from the mixed signal, thus more accurately achieving under determined blind source separation. The effectiveness of the proposed algorithm is verified by simulation experiments.