Research on the methods for electroencephalogram feature extraction based on blind source separation.
- Author:
Jiang WANG
;
Huiyuan ZHANG
;
Lei WANG
;
Guizhi XU
- Publication Type:Journal Article
- MeSH:
Algorithms;
Artifacts;
Brain;
physiology;
Brain-Computer Interfaces;
Electroencephalography;
Foot;
Hand;
Humans;
Imagination;
Movement;
Signal Processing, Computer-Assisted;
Tongue
- From:
Journal of Biomedical Engineering
2014;31(6):1195-1201
- CountryChina
- Language:Chinese
-
Abstract:
In the present investigation, we studied four methods of blind source separation/independent component analysis (BSS/ICA), AMUSE, SOBI, JADE, and FastICA. We did the feature extraction of electroencephalogram (EEG) signals of brain computer interface (BCI) for classifying spontaneous mental activities, which contained four mental tasks including imagination of left hand, right hand, foot and tongue movement. Different methods of extract physiological components were studied and achieved good performance. Then, three combined methods of SOBI and FastICA for extraction of EEG features of motor imagery were proposed. The results showed that combining of SOBI and ICA could not only reduce various artifacts and noise but also localize useful source and improve accuracy of BCI. It would improve further study of physiological mechanisms of motor imagery.