A hybrid approach to time series analysis and spectral estimation

This paper addresses the signal modeling problem in colored noise. As contrasted to the reported literatures in which the modeling of a non-Gaussian ARMA signal corrupted by Gaussian noise is studied, this paper focus on the AR modeling of a Gaussian signal embedded in non-Gaussian ARMA noise. The authors show that after prefiltering the output data via the AR polynomial of the non-Gaussian noise model, a new special higher-order Yule-Walker equation which is based on the correlation of the filtered output process can be used to estimate the parameters of the AR Gaussian signal. Simulation examples are presented to demonstrate the effectiveness of the new approach.