A Suboptimum Maximum Likelihood Approach to Parametric Signal Analysis

Abstract

A computationally efficient approach to stochastic ARMA modeling of wide-sense stationary signals is proposed. The discrete estimator minimizes a modified version of the likelihood function by using exclusively linear techniques, and thus circumventing the high computational complexity of the Maximum Likelihood (ML) method. The proposed approach is thus easy to implement, requires no explicit second order statistical information, and is shown to produce high quality estimates at a very modest computational cost. A recursive version of the algorithm, suitable for on-line implementation, is also developed, and, modeling strategy issues discussed. The effectiveness of the proposed approach is finally established through numerical simulations and comparisons with other suboptimum schemes.