EEG classification of word perception using common spatial pattern filter

The purpose of this study is to classify perceptual electroencephalography (EEG) data of words into appropriate class. We recorded EEG data from six native Korean speakers at Gwangju Institute of Science and Technology. Three words (/Lemon/, /Mother/, and /Toilet/) were chosen for stimuli. We applied IIR bandpass filter for extracting alpha bands activities from raw EEG data and common spatial pattern filter to enhance classification performance. We used pairwise classification method and mean classification rates corresponding to /Lemon/ vs /Mother/, /Lemon/ vs /Toilet/, and /Mother/ vs /Toilet/ were 54.31 ± 4.31, 59.66 ± 3.11, and 59.88 ± 5.70 respectively for all the subjects.