Research On Algorithm Of Compressed Sensing In Wireless Sensor Networks

Wireless sensor network (WSN) consists of a large number of sensor nodes withlimited computing and communication capacity, which has been mainly applied in thefield of monitoring for data acquisition and transmission. Compressed sensing (CS)theory breaks through the Nyquist sampling in traditional signal processing andcompression and sampling will be done simultaneously, so redundancy reduced at thesame time. It is effective to address the bottleneck problems of information acquisitionand transmission under the limited energy resources of WSN. CS extracts importantinformation of measurements based on sparse representation, using the measurementmatrix for dimension reduction. Finally, original signal will be restored on the terminalwith strong capability of computing. It is remarkable to improve the performance ofWSN and prolong the network life.Research applies the theory to the monitoring network of a volcano with lots ofwireless sensors deployed around to deal with the real-time data acquisition which canbe monitoring and management. Due to the limit energy and weak computationcapability of sensor nodes, as well as the strict requirements of the algorithm inmonitoring system, overflow of the storage in a sensor and the consuming computingtime must be solved in the actual data acquisition. In order to address the demand oflarge storage space of Gaussian random matrix, sparse measurement matrix suit forresource-constrained is generated with lower time complexity and space complexity.Using binary and sparse random matrix, the multiplication of measurement matrix andsignal in CS can be translated into the addition of different components of the signal.They are better than the others in the aspects of forming speed and storage capacityrequired. Theoretical proof of the Restricted Isometry Principle (RIP) of matrixes isshown. Simulation and actual test results show that the improved measurement matrixmemory can generally satisfy the small wireless sensor nodes. A CS reconstructionalgorithm by combining wavelet tree model and Iterative Hard Thresholding (IHT) ispresented simultaneously, and the convergence of the algorithm is proved theoretically.The simulation results show that compared with many kinds of popular reconstructionalgorithms, the improved algorithm has lower error bound and better robustness.After that, the proposed CS theory also applies to a distributed wireless sensornetwork. A joint sparse model based on modified measurement matrix and reconstruction algorithm is put forward. Simulation verifies its feasibility, andcomparison between joint and separate and application of a variety of algorithms injoint sparse model (JSM) is verified. The results show that compared with the separatereconstruction, JSM based on distributed compressed sensing (DCS) has a highlyaccurate reconstruction, meantime, the lower measurements can also ensurereconstruction precisely.