Information fusion estimation of noise statistics for multisensor systems
Yuan Gao, Weiling Wang, Zili Deng · 2009
For multisensor linear discrete time invariant system with unknown noise statistics and correlated noises, by the correlation method, the online local estimators of noise variances, correlated matrices and cross covariances can be obtained by solving the different partial correlated function matrix equations. The information fusion noise statistics estimators are presented by averaging the local estimators of noise statistics. Based on the ergodicity of the sample correlated function, it is proved the local and fused estimators of noise statistics are strong consistent, i.e. they converge to corresponding true values with probability one. They can be applied to design the self tuning information fusion filters. A simulation example of three-sensor system with correlated noises shows the effectiveness of the fused estimation.