A Two-stage ARMAX Identification Approach Based on Bias-eliminated Least Squares and Parameter Relationship between MA Process and Its Inverse

In this paper, a two-stage approach is proposed for the parameter identification of autoregressive moving average with exogenous (ARMAX) variable model. The proposed approach identifies the autoregressive part with exogenous variable (ARX) by a bias-eliminated least squares method, and the moving average (MA) part by utilizing the parameter relationship between MA process and its inverse. Finally, the noise variance can be computed by using the identified MA parameters. Numerical simulations validate the effectiveness of the proposed approach.