Eigenstructure based broadband source location estimation

In this paper, a high resolution eigenstructure based approach to broadband Source Location Estimation (SLE) is introduced. The approach utilizes a source low rank model, termed the source representation space. New SLE procedures are suggested in which a location contour is generated by projecting, for each proposed source location, orthonormal basis vectors for the source representation space onto either the signal or noise-only subspace of a broadband data spatio/temporal covariance matrix. Conditions for the existence of a noise-only subspace are identified. The new approach, which does not require estimates of narrowband spatial covariance matrices, has application to the limited temporal sample broadband SLE problem.