Distributed Compressed Sensing And Its Application In Coalmine Monitoring Source Coding

The popularization and application of Internet of things(IOT)in coalmine face the problems of limited network bandwidth,serious transmission interference,limited energy and limited computing capacity of wireless sensor node.Efficient,reliable and low computational complexity source coding algorithm is a key technical problem in coalmine IOT.Distributed Compressed Sensing(DCS)not only can use intra-signal correlation but also inter-signal correlation to decode,can transfer computational complexity from coder to decoder,and can get better anti-noise property by increasing measurement number appropriately.These characters make it especially suitable for coalmine monitoring source coding.Therefore,this thesis researches DCS and key technologies of applying it on coalmine monitoring source coding.Focusing on key technologies of coalmine monitoring source coding,following studies are carried out:Joint sparsity is a prerequisite of using DCS.Because coalmine gas concentration signal and coalmine image are typical coalmine monitoring sources,their joint sparsities are first studied.Research results show that they have good sparsity on discrete fourier basis.This proves that it is feasible to code coalmine monitoring source with DCS.Aiming at the problem that Gaussian random measurement matrix has uncertain performance and needed be transmitted,and dense measurement matrix has high coding complexity,deterministic sparse measurement matrix is studied.The incidence matrix of Blanced Incomplete Block Design(BIBD)proposed by Bose is defined as infrastructure framework.By embedding Discrete Fourier matrix into infrastructure framework,four types of deterministic sparse measurement matrices are constructed.They are called Bose-fourier measurement matrices.Experimental results show that,in the absence of noise and noisy conditions,Bose-fourier measurement matrix performs well on random signal coding,coalmine monitoring source(gas concentration and image)coding.Utilizing the characteristic that binary pseudo-random sequence has both randomness and good irrelevance,deterministic measurement matrix is constructed by using a good binary pseudo-random sequence-GMW(Gordon-Mills-Welch)sequence.A temporary matrix is constructed by taking GMW sequence as initial atom and other atoms cycle shifted from GMW sequence.The GMW deterministic measurement matrix is constructed by randomly choosing some rows form the temporary matrix.Experimental results show that,in the absence of noise and noisy conditions,GMW measurement matrix performs well in random signal coding,coalmine monitoring source(gas concentration and image)coding.Aiming at the problem that greedy pursuit decoding algorithms easily fall into local optimal solution,inspired by the global search ability of Grey Wolf Optimization(GWO),a joint recovery algorithm based on GWO is proposed.Solutions are simulated as grey wolves' positions.They are initialized by usingalgorithm,revaluated by using the backtracking mechanism of subspace pursuit and updated by using the idea of GWO.Inheriting above algorithms' advantages,DCS-GWO has the ability of skipping out local optimum position and searching global optimum solution.Simulation results illustrate that DCS-GWO performs well in random signal coding,coalmine monitoring source(gas concentration and image)coding.Simultaneous Hybrid Orthogonal Forward-backward Pursuiting(SHOFBP)algorithm is proposed by combining Simultaneous Orthogonal Matching Pursuit(SOMP)and Simultaneous Forward-backward Pursuit(SFBP).SHOFBP utilizes SOMP to select atoms one by one in the forward step,and utilizes the backtracking mechanism of SFBP to revaluate the selected atoms.Compared with SOMP and SFBP,SHOFBP selects atoms more accurately.Experimental results show that SHOFBP performs well in random signal coding,coalmine monitoring source(gas concentration and image)coding.