A new joint diagonalization algorithm with application in blind source separation

In this letter, we present a new nonorthogonal algorithm for joint diagonalization of a set of symmetric matrices. The algorithm alternates between updates of individual demixing matrix rows, and the update of each row is transferred to solving the eigenvector problem. By using some blind source separation simulations, we show that the algorithm obviously obtains an improved performance when the signal-to-noise ratio of the observed signals is relatively low.