An Alternative Method for ARMA(p, q) Model Characterization of Multipath Fading Channels

An alternative ARMA model parameterization method for approximating the Clarke's correlation statistics is proposed. Instead of using the standard autocorrelation function in the estimation process it considers the reflection coefficients sequence, an alternative second order statistic. It is shown that the higher the order of the ARMA(p, p-1) model, the more ill-conditioned the problem becomes. The proposed method includes an automatic ill-conditioning regularization method which does not require a priori knowledge of the regularization parameter as is the case with other methods. Two ARMA(p, q) models of different orders are obtained and discussed, in particular an ARMA(12,11) model and an ARMA(14,13) model. Both models provide very good approximations to the Clarke's autocorrelation function and they are compared with other proposed ARMA(p, q) models from the literature.