APPROXIMATE MAXIMUM LIKELIHOOD ARMA ESTIMATOR BASED ON THE POWER CEPSTRUM.

An approximate maximum-likelihood estimator is derived for ARMA (autoregressive moving-average) processes and is shown to correspond to least-squares fitting of the estimated cepstrum of the process by the model cepstrum. Experiments with several simple ARMA (1,1) processes show that the pole estimates are essentially unbiased, with variances that exceed the Cramer-Rao bound by a factor of only about two. The performance for the zero is similar if the pole and zero are not far apart, but otherwise the zero estimate is biased and has a large variance for this and all other methods tested.