Topographic Independent Component Analysis
Neural Computation
Nonparametric Supervised Learning by Linear Interpolation with Maximum Entropy
IEEE Transactions on Pattern Analysis and Machine Intelligence
Linear State-Space Models for Blind Source Separation
The Journal of Machine Learning Research
Hi-index | 0.00 |
Infomax algorithm is one of the main strategies in blind source separation. The principle and improvement of the algorithm are introduced firstly in this paper. Nineteen-channel Electroencephalograms (EEGs) which include electromyogram, eye-movement and some other artifacts were decomposed by using this algorithm. Afterwards, three kinds of nonlinear parameters were calculated for all the independent components, and artifact components can be identified automatically by threshold settings. Finally, putting all the artifact components into zero, and projecting the other components to the scalp electrodes, then the purer Electroencephalograms can be gained. The study shows that the various artifacts can be separated from the EEGs successfully with the use of adaptive Infomax algorithm and removal of artifacts can be realized by signal reconstruction. Adaptive Infomax algorithm is a potential tool in removal of artifacts in physiological signal.