ANALYSE SPECTRALE PAR DECOMPOSITION RECURSIVE EN SOUS-ESPACES PROPRES VIA LES COEFFICIENTS DE REFLEXION

A new recursive eigendecomposition algorithm of Complex Hermitian Tœplitz matrices is studied. Based on Trench's inversion of Tœplitz matrices from their autoregressive analysis, we have developed a fast recursive iterative algorithm that takes into account the rank-one modification of successive order Toep litz matrices. To speed up the computational time and to increase numerical stability of this ill-c onditioned eigendecomposition in case of very short data records analysis, we have extended this method by introducing reflection coefficients via Levinson equation, that could be easily regularized. Finally, we have developed a new regularized detector using log-likelihood ratio from reflection coefficients in place of eigenvalues previously used. Resume Nous proposons un nouvel algorithme de calcul des elements propres d'une matrice Tœplitz Hermitienne Complexe, recursivement sur l'odre de cette matrice, en tenant compte de sa structure particuliere, qui transparait a travers son rang de deplacement, comme cela est fait dans l'algorithme d'inversion de Trench. Il est possible d'accelerer l'algorithme en utilisant l'equation de Levinson et en faisant apparaitre les coefficients de reflexion. Cet algorithme est rendu robuste en cas de peu d'echantillons de donnees analyses, en utilisant des coefficients de reflexion regularises. Nous proposons egalement un nouveau test statistique robuste base sur le rapport des log-vraisemblances pour verifier l'egalite des plus petites valeurs propres, a partir egalement des coefficients de reflexion qu'il est aussi possible de regulariser pour des jeux de donnees courts.

[1]  C. Waternaux Principal components in the nonnormal case: The test of equality of Q roots , 1984 .

[2]  G. Bienvenu,et al.  Optimality of high resolution array processing using the eigensystem approach , 1983 .

[3]  M. Morf,et al.  Displacement ranks of matrices and linear equations , 1979 .

[4]  Frédéric Barbaresco Recursive eigendecomposition via autoregressive analysis and ago-antagonistic regularization , 1997, 1997 IEEE International Conference on Acoustics, Speech, and Signal Processing.

[5]  Pierre Comon,et al.  Displacement rank of pseudo-inverses , 1992, [Proceedings] ICASSP-92: 1992 IEEE International Conference on Acoustics, Speech, and Signal Processing.

[6]  James A. Cadzow,et al.  Recursive Eigenspace Decomposition, RISE, and Applications , 1994 .

[7]  C. Gueguen,et al.  An extended displacement operator for weakly structured covariance matrices , 1992, [Proceedings] ICASSP-92: 1992 IEEE International Conference on Acoustics, Speech, and Signal Processing.

[8]  G. Golub,et al.  Tracking a few extreme singular values and vectors in signal processing , 1990, Proc. IEEE.