SAUD, un algorithme d'ICA par deflation semi-algebrique

We propose in this paper a new ICA method, namely SAUD, based on a Semi-Algebraic Unitary Deflation procedure. SAUD is able to identify one by one the independent components avoiding the drawbacks of an adaptive solution (slow convergence, etc.). Computer results, representative of the biomedical context (EEG), show the good behavior of SAUD compared with other algorithms such as COM2, DEFA and FastICA.

[1]  Ben H. Jansen,et al.  Electroencephalogram and visual evoked potential generation in a mathematical model of coupled cortical columns , 1995, Biological Cybernetics.

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

[3]  Nathalie Delfosse,et al.  Adaptive blind separation of independent sources: A deflation approach , 1995, Signal Process..

[4]  Terrence J. Sejnowski,et al.  Independent Component Analysis Using an Extended Infomax Algorithm for Mixed Subgaussian and Supergaussian Sources , 1999, Neural Computation.

[5]  Laurent Albera,et al.  Brain Source Localization Using a Fourth-Order Deflation Scheme , 2008, IEEE Transactions on Biomedical Engineering.

[6]  Jutten,et al.  1 - Une solution neuromimétique au problème de séparation de sources , 1988 .

[7]  E. Oja,et al.  Independent Component Analysis , 2013 .

[8]  Phillip A. Regalia,et al.  An adaptive unit norm filter with applications to signal analysis and Karhunen-Loeve transformations , 1990 .

[9]  Pierre Comon,et al.  How fast is FastICA? , 2006, 2006 14th European Signal Processing Conference.

[10]  P. Comon,et al.  ICAR: a tool for blind source separation using fourth-order statistics only , 2005, IEEE Transactions on Signal Processing.

[11]  Lotfi Senhadji,et al.  Blind source separation for ambulatory sleep recording , 2006, IEEE Transactions on Information Technology in Biomedicine.

[12]  Comon.  5 - Analyse en composantes indépendantes et identification aveugle , 1990 .