Multichannel ARMA Signal Kalman Filter based on Covariance Intersection

Abstract

This paper is concerned with the multichannel autoregressive moving average (ARMA) signal system with colored observation noises. An ARMA signal Kalman filter based on the covariance intersection method is proposed, where the fused estimates are obtained by the local Kalman filter with unknown cross-covariance. The advantage of the proposed method is that the computation of cross-covariance is avoided, and the computational burden is significantly reduced. Under the unbiased linear minimum variance criterion, a distributed Kalman filter with matrix weights is also presented. The accuracy comparison of the covariance intersection fuser with the distributed weighted fusion Kalman filter and the local Kalman filter is given. A two-sensor ARMA signal simulation results validate the proposed Kalman filtering algorithm.