A Research On Compression Algorithm For Wireless Sensor Networks Based On Compressed Sensing

In recent years wireless sensor network (WSN) technology has got the wideattention from scholars as one of the core technology in the Internet of Things. Thestudy found that the life of Wireless Sensor Networks mainly depends on the amount ofdata transmitted by the routing node. Data compression to reduce the transmission ofinformation the amount of data can effectively prolong the network lifetime and extendrange of applications of wireless sensor networks.Aim at these issues, we introducecompressed sensing methods to wireless sensor network which compressing data andreducing the amount of data transferred effectively. We introduce distributedcompressed sensing to enhance compression. In order to introduce compressed sensingto wireless sensor network, we completed the following work.Analyze the sparsity of the monitored signals of the wireless sensor networks.Achieve independent compression and reconstruction of the single signal. In order toadapt to the application environment for wireless sensor networks, we need research thecompressed sensing observation matrix and reconstruction algorithm. This paper studiesthe impact on the accuracy of signal reconstruction and data compression performanceof the Gaussian measurement matrix, Bernoulli measurement matrix, partial orthogonalmeasurement matrix and partial Hadamard measurement matrix. Study the theperformance of matching pursuit algorithm, orthogonal matching pursuit algorithm andgradient pursuit algorithm. Determine our observation matrix and reconstructionalgorithm in practical terms by the study of these two core issues.We need to determine the number of iterations of reconstruction algorithm. Weused to determine it by empirical estimating. This approach is inaccurate, especially inthe use of some random measurement matrix. The randomness of the matrix will leadthe number of iterations to a floating number. We have selected the concept of residualupdate to decide dynamically whether to stop the iterative computation.Study the joint sparse model of distributed compressed sensing. Use theFourier-base for the projection base. We find that the public parts of the signals aremostly low-frequency and slowly varying signals. So we change the observation matrixof the joint sparse model, to avoid the reconstruction of the part of the high frequencycomponents. Analysis the algorithm to confirm the application scenarios of this method.We complete the acquisition and collection of the signals through the network. Weverify the above four ideas through the practical applications of wireless sensor networkexperiment. The experiment showed that the application of the algorithm reduces theamount of data transfered by network and the overall network power consumption.