The Reconstruction Algorithms Research Of Distributed Compressed Sensing

Compressed sensing proposed in recent years is a novel sampling theory, which isbased on the sparsity of signal. The theory has been proved that it can reconstruct theoriginal signal using less than the Nyquist sampling rate, and distributed compressedsensing established on the basis of compressed sensing can exploit both intra-and inter-signal correlation structures, so it can further reduce the sampling points required foraccurate reconstruction, it provides a new idea for multi-sensor data fusion.This paper, based on the problem of recononstructing jointly multi-signals ensembles,researches the reconstruction algorithms of distributed compressed sensing.First, to take advantage of the mixed support-set model, this paper proposes jointlook ahead orthogonal matching pursuit algorithm. Firstly, on the basis of the look aheadorthogonal matching pursuit algorithm, the initial support-set and residual are modified sothat it can use an estimate of the common support-set as an initial support-set. Then in allof the support-sets estimated by each sensor, the supports which have the highestfrequency are put into the support-set, and on this basis, the innovation components arereconstructed by the improved look ahead orthogonal matching pursuit algorithm. Theexperimental results show the effectiveness of the proposed algorithm.Second, to solve the multiple measurement vector problem, this paper proposesiterative fusion of greedy pursuit for multiple measurement vector. Firstly, on the basis ofthe forward backward pursuit algorithm, an algorithm called forward backward pursuit formultiple measurement vector is proposed by changing the method for atoms selecting;Then a strategy of fusion is introduced and in each iteration, the estimated support-set isdetermined by fusing the estimated information from orthogonal matching pursuit formultiple measurement vector algorithm and forward backward pursuit for multiplemeasurement vector algorithm; Finally, the experimental results show the effectiveness ofthe proposed algorithm.Third, Aiming at the disadvantage of the high complexity and ignoring signal’sstructural sparsity in A*orthogonal matching pursuit algorithm, a block A*orthogonal matching pursuit for multiple measurement vector algorithm is proposed. In the proposedalgorithm, the single atom is replaced by a block that is composed of several atoms and anew node is selected by projecting all blocks onto the residual matrix and selecting theblock with the smallest projection error, furthermore, the sparsity is replaced by themaximum length of all the paths on the search tree when calculating the path cost. Usingthis algorithm in the experiment section, the temperature signals which are measured bysensors in the adjacent region are jointly reconstructed perfectly.