System identification through first and second order information from the periodogram

The use of first- and second-order information in the characterization of linear systems is considered. Consideration is given to the case when this information is not available but some samples are given of a random process which are the result of filtering white noise through the system. The authors examine an approach which, starting from one estimate of the autocorrelation function, gives rise to an ARMA (autoregressive moving-average) model for the system considered. The derivation of the model is achieved from an optimization point of view.<>