基于互交替投影的块稀疏正交匹配追踪算法

针对块稀疏信号,理论分析和实验验证均表明算法精确重构的充分条件与矩阵块相关性和子相关性有关。在此基础上,提出了一种基于互交替投影的块稀疏正交匹配追踪算法(mutual alternating projection-block or-thogonal matching pursuit,MAP-BOMP)。该算法利用互交替投影方法不断构造新的测量矩阵和感知矩阵,使得矩阵块相关性和子相关性都很小,从而提高重构概率,并给出明确的算法收敛条件,降低了计算复杂度。通过与大多数已有块稀疏信号重构算法进行实验仿真对比,该算法在重构效果和重构速度上均优于其他算法。

New block orthogonal matching pursuit algorithm using mutual alternating projection method

As to block sparse signals, both theoretical analysis and experimental verification showed that sufficient condition for precise reconstruction is both block-coherence and sub-coherence of dictionary must be very small. This paper introduced a new block orthogonal matching pursuit algorithm using mutual alternating projection method MAP-BOMP. By exploiting the mutual alternating projection, the algorithm proposed to design the new measurement dictionary and sensing dictionary constantly for reducing the block-coherence and sub-coherence. The algorithm also gave the specific convergence conditions to reduce the complexity. Simulation results demonstrate this algorithm can provide more significant improvement for the recovery performance and speed than other existing algorithm.