ARMA model maximum entropy power spectral estimation

In this paper we show that the appropriate selection of constraints in the variational formulation of spectral estimation leads to new models and more freedom for the designer. We also illustrate the relationships between the set of constraints and the underlying model in the procedure. The rest of the paper concentrates on obtaining maximum entropy ARMA models for spectral estimation, using cepstral constraints and correlation constraints simultaneously. The nonlinearity that these kinds of constraints introduce is avoided by a simple linearization, that provides an estimator which is easily implemented. Finally, some examples are given to illustrate the performance of the proposed method.