Title :
A new cumulant-based criterion for iterative Blind Source Separation
Author :
Martin-Clemente, Ruben ; Acha, Jose I.
Author_Institution :
Area de Teor. de la Senal y Comun., Univ. of Sevilla, Sevilla, Spain
Abstract :
This contribution introduces a new iterative algorithm for Blind Source Separation based on the cancellation of some second order derivatives of the output cross-cumulants. It is proven that the estimation of the separating matrix can be achieved by solving linear equations repeatedly. A key feature is thus that the algorithm requires a low computational effort. Simulations show the effectiveness of the approach.
Keywords :
blind source separation; higher order statistics; iterative methods; matrix algebra; BSS; blind source separation; cumulant-based criterion; iterative algorithm; linear equations; low computational effort; second order derivatives cancellation; separating matrix estimation; Computer simulation; Cost function; Equations; Handheld computers; Mathematical model;
Conference_Titel :
Signal Processing Conference, 2002 11th European
Conference_Location :
Toulouse