Abstract: Using the Kalman filtering method, based on white noise estimation theory, under the linear minimum variance information fusion criterion, two-sensor information fusion steady-state optimal Wiener filter, smoother and predictor are presented for the multichannel Auto-Regressive Moving Average（ARMA） signals, where the optimal weighting matrices and minimum fused error variance matrix are given. Compared with the single sensor case, the accuracy of the filter is improved. A simulation example of a radar tracking system shows its effectiveness.