Reconstruction algorithm of chaotic signal based on generalized likelihood ratio threshold-decision

Abstract

Blind signal reconstruction in sensor arrays is usually a highly nonlinear and non-Gaussian problem, and nonlinear filtering is an effective way to realize state estimation from available observations. Developing the processing problem of blind signal in wireless sensor networks (WSNs) will greatly extend the application scope. Meanwhile, it also meets great challenges such as energy and bandwidth constrained. For solving the constrained problem in WSNs, the observed signals must be quantified before sending to the fusion center, which makes the overall noise unable to be modeled accurately by simple probabilistic model. To study the reconstruction issue of chaotic signal with unknown statistics in WSNs, a reconstructed method of chaotic signal based on a cost reference particle filter (CRPF) is proposed in this paper. The cost recerence cubature particle filter (CRCPF) algorithm adopts cubature-point transformation to enhance the accuracy of prediction particles, and cost-risk functions are defined to complete particle propagation. The effectiveness of proposed CRCPF algorithm is verified in the sensor network with a fusion center. Moreover, a generalized likelihood ratio functionis obtained by the cost increment of local reconstructed signals in the cluster-based sensor network topology model, which is used to reduce the network energy consumption by selecting working nodes. Simulation results show that compared with CPF and CRPF, the proposed algorithm CRCPF attains good performance in a WSN with unknown noise statistics. Meanwhile, the CRCPF algorithm realizes the compromise between energy consumption and reconstruction accuracy simultaneously, which indicates that the proposed CRCPF algorithm has the potential to extend other application scope.