Fast fixed-point algorithm for blind separation of nonlinear autocorrelation and non-Gaussian sources
Blind source separation (BSS) problem is often solved by using only one statistical property of original sources. In this work, a method combines non-Gaussianity and nonlinear autocorrelation for the BSS problem, which extends the previous BSS situation, is presented.We propose a fast fixed-point al...
Saved in:
Published in | 2011 3rd International Conference on Awareness Science and Technology (iCAST) pp. 40 - 45 |
---|---|
Main Authors | , , , |
Format | Conference Proceeding |
Language | English |
Published |
IEEE
01.09.2011
|
Subjects | |
Online Access | Get full text |
Cover
Loading…
Summary: | Blind source separation (BSS) problem is often solved by using only one statistical property of original sources. In this work, a method combines non-Gaussianity and nonlinear autocorrelation for the BSS problem, which extends the previous BSS situation, is presented.We propose a fast fixed-point algorithm for BSS with nonlinear autocorrelation and non-Gaussianity in this paper. Our algorithm obtained here does not need choose any learning rate. We study its convergence property and show that its convergence speed is at least quadratic. Computer simulations for square temporal autocorrelation and non-Gaussian sources, including sub-Gaussian and super-Gaussian sources, illustrate the efficiency of the proposed approach. |
---|---|
ISBN: | 1457708876 9781457708879 |
ISSN: | 2325-5986 2325-5994 |
DOI: | 10.1109/ICAwST.2011.6163093 |