Unsupervised fuzzy c-means clustering for motor imagery EEG recognition

In this study, an electroencephalogram (EEG) recognition system is proposed on single-trial motor imagery (MI) data. Fuzzy c-means (FCM) clustering is used for the unsupervised recognition of left and right MI data by combining with selected active segments and multiresolution fractal features. Active segment selection is used to detect active segments situated at most discriminable areas in the time-frequency domain. The multiresolution fractal features are then extracted by using modified fractal dimension from wavelet data. Finally, FCM clustering is used as the discriminant of MI features. The FCM clustering is an adaptive approach suitable for the clustering of non-stationary biomedical signals. Compared with several popular supervised classifiers, FCM clustering provides a potential for BCI application.