Some signal processing methods based on covariance estimates of signal and noise

Three methods for detection and estimation with sensor arrays are described, together with the procedures developed for studying the methods by computer simulation. Two of them are generalized versions of the classical square law detector and Capron's detector. The third is based on a detector using the generalized eigenvalues and eigenvectors of the signal covariance matrice. Only the square law detector is not a so called high resolution method. Results of numerical simulations show the directional characteristics of the detectors for exact covariance matrices, and for covariance matrice estimates, produced by the update algorithm using computer generated signal and noise samples. The performance of all three detectors appears to be better with the fast algorithm for approximative rank-one updates of the covariance matrice, than with exact rank-one updates.