Title :
Blind identifiability of class of nonlinear instantaneous ICA models
Author :
Eriksson, Jan ; Koivunen, Visa
Author_Institution :
Signal Process. Lab., Helsinki Univ. of Technol., Helsinki, Finland
Abstract :
The identifiability/separability of nonlinear instantaneous ICA models is considered. The identifiability proof is constructed for the class of nonlinearities satisfying addition theorem. Addition theorem covers wide variety of nonlinear mixing systems of engineering interest. An algorithm for separating such nonlinear mixtures is presented and the feasibility of the approach is demonstrated.
Keywords :
blind source separation; image processing; independent component analysis; blind identifiability; blind separability; independent component analysis; nonlinear instantaneous ICA model; nonlinear mixing system; Abstracts; Reliability; Signal processing;
Conference_Titel :
Signal Processing Conference, 2002 11th European
Conference_Location :
Toulouse
