Common frequency pattern for music preference identification using frontal EEG

In this paper, we investigate the use of 2-channel frontal EEG signal to classify two music preferences: like and dislike. The hypothesis for this investigation is that the frontal EEG signal contains sufficient information on the mental state of a subject for discriminating the preference of music of the subject. An experiment is performed to collect 2-channel frontal EEG data from 12 subjects by playing various types of music pieces and asking whether they like or dislike the music in order to obtain the true labels of their music preferences. We then propose a frequency band optimization method called common frequency pattern (CFP) for feature extraction and Linear SVM for classification to identify the music preference of the subjects from the 2-channel frontal EEG. The results of using the proposed method yield an average classification accuracy of 74.77% for a trial length of 30 s over the 12 subjects. Hence the experimental results show evidence that frontal EEG signal contains sufficient information to discriminate preference of music. Furthermore, the frequency band optimization results indicate that gamma band is essential for EEG-based music preference identification.