A Nonlinear Dynamical Chaotic Signal Reconstruction Method In Wireless Sensor Networks With Unknown Statistics

Blind signal processing (BSP) in wireless sensor networks (WSNs) often encounters circumstances where state estimation is modeled by discrete-time dynamical random systems, and observed signals are corrupted by noise of unknown statistics. Cost reference particle filter (CRPF) is an efficient recently proposed methodology for state estimation of nonlinear dynamical systems of unknown statistics. By combining the cubature-points rules with the CRPF algorithm, this paper proposes a new cost reference cubature particle filter (CRCPF) for chaotic signal reconstruction in a WSN system with unknown noise statistics. Computer simulations are used to demonstrate the effectiveness and robustness of the CRCPF algorithm when compared with cubature Kalman filter (CKF) and cubature particle filter (CPF). The results show that the CRCPF algorithm attains better performance than the other two methods for mixture noise of unknown statistics.