When multiple correlated data channels are simultaneously processed, multichannel-autoregressive (M-AR) processes may be combined with optimal filtering such as Kalman or H<sub>∞</sub> for prediction or estimation from noisy observations. However, the estimation of the M-AR parameters from noisy observations is a key issue to be addressed. Off-line or iterative approaches have been proposed recently, but their computational costs are high or some of them may diverge. Using on-line approaches such as EKF and SPKF is of interest, but the size of the state vector to be estimated is quite high. To reduce this size and the resulting computational costs, we suggest using dual optimal filters. In this paper, we study the relevance of cross-coupled Kalman filters and cross-coupled H<sub>∞</sub> filters. The comparative simulation study we carry out shows that our approach corresponds to a compromise between computational cost and performances in terms of pole estimation accuracy.