Parametric Spectral Estimation: A Fast Rational Model Approach

A novel approach to fast ARMA spectral estimation is proposed. The proposed approach is based on a correlation version of the recently introduced Suboptimum Maximum Likelihood technique, and significantly reduces the computational and memory storage requirements, while maintaining the high estimation accuracy. The consistency of the proposed estimator is established, and a modified version that overcomes the algorithmic instability problem associated with most available techniques, developed. The effectiveness and accuracy of the proposed approach with both long and short data records is finally established through numerical simulations and comparisons with the Blackman-Tukey and the AR spectral analysis methods.