An approach to direction finding based on a subspace perturbation expansion

This paper describes a method for estimating directions of arrival from sensor-array data. The estimates are obtained from minimizing a subspace-fitting cost function. This cost function is optimally weighted using statistical information provided by a subspace perturbation expansion. The calculations for scenarios involving very closely spaced sources suffers from the need to invert ill conditioned matrices. This problem is overcome by a reparameterization of both the cost function and its Jacobian based on the concept of a limiting subspace. The paper includes a challenging simulation example involving multiple moving sources.