High-resolution analysis of non-stationary data ensembles

Abstract

Estimation of an unknown parameter of a random data ensemble is usually done by maximizing a certain power estimator by varying the unknown parameter. This method is convenient where the data are a nonlinear function of the parameter. It is shown for a variety of power estimators that high resolution is always based on the orthogonality between the observed data and a weighting vector that contains the parameter. Some examples illustrate the use of these power estimators to locate sources by means of array antennas.