Compressed Sensing For Data Fusion In Wireless Sensor Network

Compressed sensing technology collectes sparse signal at the rate far below Nyquistrate, which provides a new method for data fusion technology in the wireless sensornetwosk. In data fusion process, it uses signal redundancy to collect and transmit. At thereceiving end, it recontructs signal by solving optimation problem, realizing the highefficiency of data collection and low loss of data transmission. At present, compressedsensing technology is widely applied in environmental monitor, medical and health, andintelligent identification and so on.This paper, based on the temporal-spatial correlation of the signal, researches theapplication of the compressed sensing in data fusion technology of wireless sensornetwork.First, to the weakness that traditional compressed sensing technology can not makefull use of the redundancy of signal, this paper points out a data aggregation techniquecombined temporal-spatial correlation and compressed sensing. This technique obtainlocal random projections based on the time correlation of signal firstly, realizing the firstcompression and then complete the second compression by using the spatial correlationand fusing routing. The decode process reconstructs signals by the iterative thresholdingalgorithm based on temporal-spatial correlation. The experimental results show that theproposed data aggregation technique reduces the loss of energy significantly.Second, to the difficulty of sparse represention for the signals of irregular network,this paper points out nonlocal spectral graph wavelet transform based on temporal-spatialcorrelation. This method constructs graph according to the postion of sensor nodes and thesimilarity of time sample between nodes, and contacts signal domain and graph Laplaciandomain, realizing the sparse representation of irregular signal. Experiment uses thetemporal-spatial correlation nonlocal spectral graph wavelet and iterative thresholdingalgorithm to reconstruct signal, the results show the effectiveness of the proposedalgorithm.Third, to the unstability of reconstruction result obtained from the data aggregation technique based on random route and the energy consumption of network transmission,this paper points out low-power compressed sensing clustering data aggregationtechnology. In order to minimize network transmission loss, algorithm uses kmeans idea toget clusters. The nodes at the same cluster use the redundancy to project repeatly. Thecluster head eventually sents a small amount of projection to the sink node. Theexperimental results show the algorithm can get stable reconstruction effect in energyefficiency.