Abstract
The identification problem is considered for an AR process without a priori knowledge of the model order from a finite set of noisy observations. Such a AR process is considered as an ARMA model. A new method to solve the present identification problem is proposed by making use of the powerful concepts of positive functions and bounded functions, based on Pillai's algorithm theory. It is shown that the deterministic theory founded on well established passive network concepts is also useful in determining order and estimating parameters for noisy AR processes.
