贪婪追踪系列算法的分析与优化

Analysis and Optimization of Greedy Pursuit Algorithms

从条件概率的角度分析了压缩感知信号重构,认为重构信号是在观测信号发生的条件下所发生的后验信号.基于信号的条件概率分析,提出一种简单而有效的算法优化方法,此方法可以提高贪婪追踪系列算法选择原子的准确性,在某种程度上可以削减感知矩阵自身相关度的影响.此外,本文概述了系列贪婪追踪算法,并应用本文提出的方法来改进它们.模拟实验表明,本文提出的方法可以改进大多数贪婪追踪算法,使算法在速度和精度上都有所提高.

The paper analyzes the reconstruction signal of compressed sensing from perspective of conditional probability, and consid- ers the reconstruction signal as mean of posterior signal given observed signal. Based on the analysis, the paper puts forward an sim- ple and efficient method toward serials greedy pursuit algorithms. This method can promote the accuracy of selection columns of per- ception matrix and reduce the influence of self-correlation of perception matrix to some extent. Meanwhile, this paper reviews most of greedy pursuit algorithms, and applies the proposed method to improving them. Simulation experiments show that the proposed meth- od can improve most greedy pursuit algorithms, whether in the reconstruction precision or in the speed.