论文信息 - Sur la diagonalisation conjointe approchée par un critère des moindres carrés

Sur la diagonalisation conjointe approchée par un critère des moindres carrés

A new algorithm for approximate joint diagonalization of a set of matrices is presented. Using the mean square criterium, without the ortogonality constraint, it is compared with an analogous algorithm for sources separation. The criterium of our algorithm is on the separating matrix while the other is on the mixing matrix. This improves significantely the rate of convergence with better performances.

S. Dégerine | Serge Degerine

[1] Abdelhamid Zaidi. Un nouvel indice de performance et sa borne inférieure , 2001 .

[2] Dinh Tuan Pham,et al. Joint Approximate Diagonalization of Positive Definite Hermitian Matrices , 2000, SIAM J. Matrix Anal. Appl..

[3] J. Cardoso,et al. Blind beamforming for non-gaussian signals , 1993 .

[4] Eric Moulines,et al. A blind source separation technique using second-order statistics , 1997, IEEE Trans. Signal Process..

[5] M. Wax,et al. A least-squares approach to joint diagonalization , 1997, IEEE Signal Processing Letters.

[6] S. Dégerine,et al. Maximum de vraisemblance exact pour la séparation aveugle d'un mélange instantané de sources autorégressives gaussiennes , 1999 .

[7] Dinh Tuan Pham,et al. Blind separation of instantaneous mixture of sources via the Gaussian mutual information criterion , 2000, 2000 10th European Signal Processing Conference.

[8] A. Ostrowski. Solution of equations in Euclidean and Banach spaces , 1973 .