Distributed Information Fusion Estimation Of Arma Signal

Signal processing has a wide application in communication and control. ARMA (autoregressive moving average) signal can be applied to many fields including time-series analysis, system modeling, deconvolution and system prediction, and so on. Signal estimation can be a special form of state estimation, such as ARMA signal being estimated as the component of a state. Since multiple sensors can supply more information in time and space, the multi-sensor information fusion technology for signal processing is wide studied and applied. In this paper, we study the fusion Wiener estimators for multi-sensor multichannel ARMA signal, the information fusion deconvolution filter for ARMA signal, and the fusion estimators for ARMA signal with measurement delays.For multi-sensor multichannel ARMA signal system, by the transform from ARMA model to state space model, the signal estimation is transferred to estimation of the component of the state, or white noise estimation and output prediction problems. Based on the optimal weighted fusion algorithms in the linear minimum variance sense, the optimal information fusion distributed Kalman estimators and Wiener estimators are given by applying estimators of white noise and output predictors. The estimation error cross-covariance matrix between any two sensors is derived.A distributed fusion deconvolution for multichannel ARMA signal is given by using the optimal fusion weighting algorithm. By the transform between ARMA model and the state model, a distributed optimal fusion deconvolution for multichannel ARMA signal in the linear minimum variance is advanced. We also derive the computational formula of the filtering error cross-covariance matrix between any two sensors.For the system with multi-step measurement delays, we transfer it to an equivalent system without measurement delays by state augmentation, and then solve it by using existing results. In addition, for the system with one-step measurement delay, we transfer it to a normal system with the correlated noise at neighbored time by a model transform, and then obtain the local estimator in the linear minimum variance sense for every sensor subsystem by applying projection theory. Then we derive the computation formula for the cross-covariance matrix between any two local estimators. At last, a distributed weighted fusion Kalman estimator is given based on the distributed optimal fusion estimation algorithm weighted by matrices.