ARMA Model Order Estimation for Coloured Gaussian Noise Based on the Eigenvalues of the Covariance Matrix
Department of Electronics Engineering, Hijjawi Faculty for Engineering Technology Yarmouk University, Irbid, Jordan
ABSTRACT: The problem of estimating the order of ARMA models has attracted much attention because of its broad application to many fields. A well-known solution to this problem is minimum eigenvalues (MEV) criterion. This method is derived from the minimum description length (MDL) criterion. The MEV method is based on the minimum eigenvalues of a family of covariance matrices from the observed data. In the MEV method, observation and/or modelling error is assumed to be zero-mean white Gaussian. In this paper, we will generalize the original results of the MEV technique to the case of coloured Gaussian noise for the second order statistics (i. e., autocorrelation).