Compression And Detection Of Data In WSN Based On Compressive Sensing

With the development of communication technology, the research on Wireless Sensor Network(WSN) has dawn great attention. Sensor nodes scattered in WSN adopt tiny powered batteries andmulti-hop transmission approach to communicate with each other, which will cause the problem oflimited energy and unbalanced load. The transmission, processing and storage of data in WSN willconsume a lot of energy, thus using compressed sensing to compress data is an effective way tosave energy. To balance load and save energy for prolonging the life of the network, we investigatedata compression and sparse reconstruction by using compressed sensing. To balance the load inWSN, we construct a new sensing matrix to save energy. Moreover, to ensure the accuracy ofreconstructing data, we apply an improved principal component analysis based on the Bayesianmodel with feedback. The main context of this thesis is as follows：First, without loss of reconstruction accuracy we construct a new sensing matrix, replaceingGaussian random matrix, by considering the combination routing choice to implement the loadbalance of WSN, which splits the nodes into small groups to transmit data, and makes a slightadjustment of the feedback coefficients. Some simulations were conducted to demonstrate theeffectiveness.Second, we use an improved principal component analysis (PCA) to compress the data inWSNs. Since the discrete cosine transform and wavelet transform are only suitable for smoothsignals and signals with spikes. In view of the signals collected in WSNs may be smooth or havespikes, and have some statistics features, we use PCA to impress the data in WSNs. Meanwhile,since the data were transported layer by layer, we use PCA at each layer to compress the data forsaving energy-consumption. Some simulations of data compression and energy-consumption weregiven to show the effectiveness using multiple-PCA and single PCA. The results show thatmultiple-PCA can improve the data compression ratio and reduce the energy consumption.Finally, we use an improved Bayesian model with feedback to reconstruct signals. Thismethod can achieve the adaptivity of systems. Comparing to BP and OMP algorithms, the givenalgorithm can improve the accuracy of signal reconstruction. We show the effectiveness by someexperimental results.