Automatic ocular artifact suppression from human operator's EEG based on a combination of independent component analysis and fuzzy c-means clustering techniques

Independent component analysis (ICA) and fuzzy c-means (FCM) clustering were adopted for automatic ocular artifact suppression from operator's electroencephalogram. Firstly, ICA was applied to the 20s data containing nine channels of EEG data and one of electrooculagram (EOG) data. Secondly, each 20s independent component (IC) was partitioned into ten 2 s epochs. And five features of each epoch were calculated, which are wavelet entropy, power in the band between 0 and 5 Hz, kurtosis, mutual information and correlation. Thirdly, the epochs were classified as either EEG or ocular artifact based on the result of FCM clustering. And then components which were recognized as ocular artifact were rejected. Clean EEG was obtained. The result shows that the method based on ICA and FCM can be applied to online automatic ocular artifact suppression from EEG.