面向混合支撑集模型的分布式压缩感知重构算法

摘要

Focusing on the mixed support-set model, a study on joint reconstruction of signals based on distributed com-pressed sensing ( DCS ) was conducted, and a joint look ahead variable stepsize orthogonal matching pursuit ( LAVSOMP) algorithm was brought forward.The joint LAVSOMP algorithm dynamically performs the adaptive ad-justment of forward parameters according to the energy difference between reconstructed signals of adjacent iterations to strike a balance between the signal reconstruction accuracy and its running time.Furthermore, a joint forward-backward variable stepsize orthogonal matching pursuit ( FBVSOMP) algorithm was put forward.The joint FBV-SOMP algorithm effectively reduces the chance of choosing non-optimal atoms and improves the signal reconstruction accuracy.The experimental results show that the joint LAVSOMP algorithm outperforms the joint look ahead orthog-onal matching pursuit algorithm fixing forward parameters in terms of reconstruction performance, and the joint FB-VSOMP algorithm can achieve the higher joint reconstruction performance than the joint LAVSOMP.%针对混合支撑集模型，研究了分布式压缩感知（ DCS）的信号联合重构，提出了一种联合向前变步长正交匹配追踪（ joint LAVSOMP）算法，该算法在信号重构过程中根据相邻次迭代重建信号的能量差，自适应地对向前参数进行动态调整，在信号重建精度与算法运行时间上取得平衡。进而，在该算法的基础上，提出了一种联合向前向后的变步长正交匹配追踪（ joint FBVSOMP）算法，该算法有效降低了原子误选的几率，提高了信号重建的精度。试验结果表明， joint LAVSOMP算法的重构性能优于向前参数固定的联合向前正交匹配追踪joint LAOMP算法，而joint FBVSOMP算法具有更高的信号联合重构性能。