For solving the issues of the signal reconstruction of nonlinear non-Gaussian signals in wireless sensor networks （WSNs）, a new signal reconstruction algorithm based on a cubature Kalman particle filter （CKPF） is ...For solving the issues of the signal reconstruction of nonlinear non-Gaussian signals in wireless sensor networks （WSNs）, a new signal reconstruction algorithm based on a cubature Kalman particle filter （CKPF） is proposed in this paper. We model the reconstruction signal first and then use the CKPF to estimate the signal. The CKPF uses a cubature Kalman filter （CKF） to generate the importance proposal distribution of the particle filter and integrates the latest observation, which can approximate the true posterior distribution better. It can improve the estimation accuracy. CKPF uses fewer cubature points than the unscented Kalman particle filter （UKPF） and has less computational overheads. Meanwhile, CKPF uses the square root of the error covariance for iterating and is more stable and accurate than the UKPF counterpart. Simulation results show that the algorithm can reconstruct the observed signals quickly and effectively, at the same time consuming less computational time and with more accuracy than the method based on UKPF.展开更多
基金supported by the National Natural Science Foundation of China(Grant No.60872123)the Joint Fund of the National Natural Science Foundation andthe Guangdong Provincial Natural Science Foundation,China(Grant No.U0835001)+2 种基金the Fundamental Research Funds for the Central Universities of Ministryof Education of China(Grant No.2012ZM0025)the South China University of Technology,Chinathe Fund for Higher-level Talent in GuangdongProvince,China(Grant No.N9101070)
文摘For solving the issues of the signal reconstruction of nonlinear non-Gaussian signals in wireless sensor networks （WSNs）, a new signal reconstruction algorithm based on a cubature Kalman particle filter （CKPF） is proposed in this paper. We model the reconstruction signal first and then use the CKPF to estimate the signal. The CKPF uses a cubature Kalman filter （CKF） to generate the importance proposal distribution of the particle filter and integrates the latest observation, which can approximate the true posterior distribution better. It can improve the estimation accuracy. CKPF uses fewer cubature points than the unscented Kalman particle filter （UKPF） and has less computational overheads. Meanwhile, CKPF uses the square root of the error covariance for iterating and is more stable and accurate than the UKPF counterpart. Simulation results show that the algorithm can reconstruct the observed signals quickly and effectively, at the same time consuming less computational time and with more accuracy than the method based on UKPF.
