GEESE (GEneralized Eigenvalues Utilizing Signal Subspace Eignevectors) - A New Technique For Direction Finding

A new technique for estimating the directions of arrival of multiple signals utilizing the generalized eigenvalues associated with certain matrices generated from the signal subspace eigenvectors is reported here. This is carried out by observing a well-known property of the signal subspace: i.e., in presence of uncorrelated and identical sensor noise, the subspace spanned by the true direction vectors coincides with the one spanned by the eigenvectors corresponding to all, except the smallest set of repeating eigenvalue of the array output covariance matrix. Further, a first-order perturbation analysis is carried out to evaluate the performance of this new scheme, when the array output cross-covariances are estimated from the data.