Improved ML cross-spectral estimation

Abstract

This work deals with the use of dedicated filters for cross-spectrum estimation. Basically, the ML cross-spectral estimator can be obtained as the natural extension of the Normalized Maximurn Likelihood procedure, reported previously by the authors, te the measurement of cross-power density for two data registers x(n) and y(nl. As an important improvement in present cross-spectrum estimation, the importance of the selection of the cross-correlation matrix estimator used as a starting point is included.

Peer Reviewed

Postprint (published version)