Circuit theoretic approach to the modeling of ARMA processes

The authors present a methodology for the autoregressive moving average (ARMA) modeling based on the circuit theoretic viewpoint. The authors discuss the convergence of the Levinson algorithm in AR modeling, and show the ineffectiveness of the Levinson algorithm in some cases where ARMA models have to be taken into consideration. Next, they show how the convergence property of the Levinson algorithm can be improved, and give a more general method for approximating the nonrational power spectral density function, by which the modeling of a stochastic process is accomplished as an ARMA model. Finally, simulation results are also shown to illustrate the proposed procedure