An Advanced Parameter Estimator of Multichannel Autoregressive Signals from Noisy Observations

Abstract

We consider the problem of estimating the multichannel autoregressive (MAR) model parameters using noisy observations. The existing improved least-squares algorithm for vector processes (ILSV) estimates both the MAR parameters and the variance-covariance matrix of the multichannel noise in an iterative manner, but it neglects the fact that the variance-covariance matrix should be symmetric. In this paper, we introduce an advanced estimator for MAR parameters and variance-covariance matrix of observation noise, which ensures the latter to be symmetric in each iterative process. In the simulations, the performance of the proposed algorithm significantly outperforms that of ILSV method.