Blind source separation of positive and partially correlated data.

Fiche publication


Date publication

septembre 2005

Auteurs

Membres identifiés du Cancéropôle Est :
Dr NUZILLARD Jean-Marc


Tous les auteurs :
Naanaa W, Nuzillard JM

Résumé

Blind source separation (BSS) consists of processing a set of observed mixed signals to separate them into a set of unobservable original components. Various approaches have been employed to solve BSS problems using the strong assumption focusing on mutually uncorrelated (or orthogonal) source signals. However, in many real-life problems, signal orthogonality is not guaranteed. This paper introduces a new approach to BSS that can be applied to nonorthogonal signals. The orthogonality requirement is replaced by a partial orthogonality and a nonnegativity constraint which are well-suited for many real-world signals. An algebraic property is then exploited to express BSS problems in terms of constrained optimization. An efficient algorithm implementing the approach is reported and applied to examples from nuclear magnetic resonance spectroscopy.

Référence

Signal Process. 2005 Sep;85(9):1711-22.