Title :
Innovation enhanced Infomax for separating linear instantaneous mixtures
Author :
Xu, YiLing ; Yang, Le ; Peng, Qicong ; Lin, Jingran
Author_Institution :
TI-DSP Lab., Univ. of Electron. Sci. & Technol. of China, Chengdu, China
Abstract :
The paper introduces an innovation process into the Infomax algorithm developed by A.J. Bell and T.J. Sejnowski (see Neural Computation, vol.7, p.1129-59, 1995). The simulation results show that while the performance of the original algorithm on super-Gaussian sources is unaffected, its performance on sub-Gaussian sources is significantly enhanced by the innovation process.
Keywords :
independent component analysis; source separation; ICA; independent component analysis; innovation enhanced Infomax algorithm; linear instantaneous mixture separation; subGaussian sources; superGaussian sources; Artificial neural networks; Blind source separation; Brain modeling; Independent component analysis; Power generation; Signal generators; Signal processing; Signal processing algorithms; Technological innovation; Unsupervised learning;
Conference_Titel :
Communications, Circuits and Systems, 2004. ICCCAS 2004. 2004 International Conference on
Print_ISBN :
0-7803-8647-7
DOI :
10.1109/ICCCAS.2004.1346351
