Separation of independent components from data mixed by several mixing matrices

Authors:
Wakako Hashimoto
Affiliations:
Graduate School of Engineering Science, Osaka University, 1-3 Machikaneyama, Toyonaka 560-8531, Osaka, Japan and Laboratory for Advanced Brain Signal Processing, Brain Science Institute, RIKEN, 2- ...
Venue:
Signal Processing - Signal processing with heavy-tailed models
Year:
2002

Citing 9
Cited 2

An information-maximization approach to blind separation and blind deconvolution

Neural Computation
Stability analysis of learning algorithms for blind source separation

Neural Networks
Natural gradient works efficiently in learning

Neural Computation
High-order contrasts for independent component analysis

Neural Computation
Unsupervised classification with non-Gaussian mixture models using ICA

Proceedings of the 1998 conference on Advances in neural information processing systems II
ICA Mixture Models for Unsupervised Classification of Non-Gaussian Classes and Automatic Context Switching in Blind Signal Separation

IEEE Transactions on Pattern Analysis and Machine Intelligence
Emergence of Phase- and Shift-Invariant Features by Decomposition of Natural Images into Independent Feature Subspaces

Neural Computation
Blind separation of mixture of independent sources through aquasi-maximum likelihood approach

IEEE Transactions on Signal Processing
Blind source separation-semiparametric statistical approach

IEEE Transactions on Signal Processing

Information-preserving hybrid data reduction based on fuzzy-rough techniques

Pattern Recognition Letters
The use of entropy minimization for the solution of blind source separation problems in image analysis

Pattern Recognition

Quantified Score

Hi-index	0.00

Visualization

Abstract

A new algorithm is proposed for the variation of independent component analysis (ICA) in which there are several mixing matrices and, for each set of independent components, one of the matrices is randomly chosen to mix the components. This method can be used to analyze a class of data generated overcompletely, and to classify data in an unsupervised manner. In the algorithm proposed by Lee et al. (IEE Trans. Pattern Anal. Mach. Intell. 22 (2000) 1078), mixing matrices were estimated by means of maximum likelihood estimation. However, if the presumed probability density function of independent components is wrong, the estimations obtained from their algorithm are not consistent. Under the same conditions, the algorithm proposed in this paper, utilizing high-order moments, can obtain consistent estimators. The effectiveness of our algorithm is verified by numerical experiments.