A generalized formulation for harmonic retrieval in correlated noise

In this paper we develop a generalized formulation for harmonic retrieval in correlated noise. Well-known spatial smoothing-based ESPRIT and the Matrix Pencil methods are important special cases of the proposed formulation. The connection between these methods is established analytically. The proposed generalized formulation is applicable to single-snapshot scenarios with arbitrary noise covariance matrix. Moreover, a numerical study is included showing that the Matrix Pencil method is robust to mismodeling of the noise covariance matrix.