Algorithms for nonorthogonal approximate joint block-diagonalization

Approximate joint block diagonalization (AJBD) of a set of matrices has applications in blind source separation, e.g., when the signal mixtures contain mutually independent subspaces of dimension higher than one. In this paper we present three novel AJBD algorithms which differ in the final target criterion to be minimized in the optimization process. Two of the algorithms extend the principle of Jacobi elementary rotations to the more general concept of matrix elementary rotations. The proposed algorithms are compared to existing state-of-the art AJBD algorithms.