COVARIANCE-CONSTRAINED MAXIMUM LIKELIHOOD METHOD.

Summary form only given. The covariance-constrained maximum-likelihood method (CCMLM) is a new estimator of the power spectrum that exhibits superior fidelity over that of alternative, high-resolution methods for certain class of spectra. The authors derive the general form of the CCMLM and discuss an iterative procedure for its implementation. Convergence properties of this procedure are presented and the issue of convergence with respect to a spectral fidelity criterion is raised. Finally, examples of the method's performance for ARMA and nonrational spectra are discussed and compared with those using alternative, high-resolution methods.