Data pre-processing for high-resolution adaptive algorithms

Abstract

The inclusion of adaptive methods in high-resolution spectral estimation algorithms is considered. The generation of a complete family of spectral estimators from the normalized maximum-likelihood method (MLM) is discussed. It is shown how the generalized power MLM can be used to generate adaptive schemes for improving resolution. The authors propose the substitution of the conventional mean-square filtering error by quadratic objectives built as inner products of the coefficient error vector of the estimator filter