Research Of The Data Processing Method Based On Compressed Sensing In WSN

The traditional data transmitting and processing method in wireless sensor networks is to send the data, which is collected by many sensor nodes as a unit, after a simple preprocessing. It will cause a large number of redundant data transmitted. As a result, the node energy consumption, the cost of network construction and the failure rate will be high. In order to solve the above problems, the researchers have done a lot of work, such as, introducing transform coding method. Although transform coding method can reduce the number of sending data, the energy consumption of the terminal nodes to calculate is large. Therefore, the researchers introduce the compressed sensing method, which transfers most of the computational burden to a coordinator with high energy and computing ability.There are two application of compressed sensing in wireless sensor networks mainly used: space and time-space method. Space method is mainly used in large-scale network. In order to make better use of compressed sensing in wireless sensor networks, researchers put the time factor into account to raise the time-space method. When time-space compressed sensing method is used in multiple hops "chain" topology network, the number of data transmitted by relay node close to the coordinator is larger than any other node’s, which leads to serious unbalanced energy consumption of each node in the network. This will cause the network life cycle short. At the same time, for the performance of all methods in a real node, the researchers did not do detailed analysis. In view of the above problems, the main research contents of this paper are as follows:1. After compressed sensing is introduced simply, the projection domain, the observation matrix and the reconstruction algorithm are choosen for the collected data. In order to analyze the performance of compressed sensing space method in wireless sensor networks, the mixed congruential method for generating random weighting coefficient and the shortest path method for routing jointly build space observation matrix. From the simulation we can see that in order to make the stochastic observation matrix, the method is only suitable for large scale network. Based on above shortcomings, time-space method that is improved from space method, is analyzed. The original data are observed by ‘0-1’ random measurement matrix, which is produced by node by physical address. Through the method of routing, measured values is sent to coordinator, which restructures the original data jointly. The theoretical analysis shows that the time-space method in the "chain type" network topology will cause node energy consumption unbalance.2. Target at the problem of the unbalance of the compressed sensing’s node energy consumption in time-space method, balanced compressed sensing method is proposed. The method equals the quantity of computation and data transmission, makes the balance of the node energy and extends lifetime of network. At the same time, the method also solves the problem of data reconstruction delay, with the ability to deal with abnormal sensing data.3. In order to explore the influence of each method on the performance of the network in the actual node, this thesis designs the hardware nodes to analyze the actual node energy consumption and set up the network model. In the multi hop network, we compare impact on the network life cycle during the two kinds of methods, namely compressed sensing methods--time-space and energy balance and the traditional method (noncompressed sensing methods)--the direct transmission and code conversion. The following conclusions are made:(1) In the multi node case, when the balanced compressed sensing method is used, life cycle of network is longer;(2) When the number of nodes is small and sparse signal rate is relatively small, performance of time-space mothed is better;(3) When the number of nodes is small and sparse signal rate is relatively large, both direct transmission and transform coding method (noncompressed sensing methods) may be chosen;(4) With the increase of node communication distance, the network life cycle of various corresponding methods will reduce and then effective method is chosen according to the number of nodes and signal sparse rate.The conclusions of this thesis can provide the basis for the practical application of the choice during the methods above.