DocumentCode :
1239652
Title :
Information theoretic versus cumulant-based contrasts for multimodal source separation
Author :
Vrins, Frédéric ; Verleysen, Michel
Author_Institution :
UCL Machine Learning Group, Univ. Catholique de Louvain, Louvain-la-Neuve, Belgium
Volume :
12
Issue :
3
fYear :
2005
fDate :
3/1/2005 12:00:00 AM
Firstpage :
190
Lastpage :
193
Abstract :
Recently, several authors have emphasized the existence of spurious maxima in usual contrast functions for source separation (e.g., the likelihood and the mutual information) when several sources have multimodal distributions. The aim of this letter is to compare the information theoretic contrasts to cumulant-based ones from the robustness to spurious maxima point of view. Even if all of them tend to measure, in some way, the same quantity, which is the output independence (or equivalently, the output non-Gaussianity), it is shown that in the case of a mixture involving two sources, the kurtosis-based contrast functions are more robust than the information theoretic ones when the source distributions are multimodal.
Keywords :
blind source separation; entropy; higher order statistics; independent component analysis; blind source separation; cumulant-based contrasts; entropy; independent component analysis; information theoretic; kurtosis-based contrast functions; multimodal distributions; spurious maxima; Covariance matrix; Entropy; Gaussian distribution; Independent component analysis; Machine learning; Mutual information; Robustness; Source separation; Blind source separation; contrast function; entropy; independent component analysis; kurtosis; multimodal sources;
fLanguage :
English
Journal_Title :
Signal Processing Letters, IEEE
Publisher :
ieee
ISSN :
1070-9908
Type :
jour
DOI :
10.1109/LSP.2004.840863
Filename :
1395937
Link To Document :
بازگشت