Overdetermined blind source separation using approximate joint diagonalization
Taiki Asamizu, Shinya Saito, Kunio Oishi, Toshihiro Furukawa · 2017
Blind separation of mixtures has been achieved by approximate joint diagonalization (AJD) approaches. This paper presents an approach for overdetermined blind source separation (BSS) using AJD. The approach is based on an alternative minimization of the indirect and direct least-squares criteria to the diagonal matrices in the first phase and to the mixing matrix in the second phase, respectively. Simulation result demonstrates that the proposed algorithm is capable for achieving better separation performance in overdetermined mixtures than in determined mixtures at reduced complexity.