Subject independent affective states classification using EEG signals

Affective states classification has become an important part of the Human-Computer Interface (HCI) study. In recent years, studies of physiological signals, such as ECG, GSR and EEG on affective expression have shown very promising results. In this study, we carried out two experiments to better understand the neurological expression of emotions through the use of EEG signals. In particular, we carried out a subject-independent affective states classification study using narrowband spectral power of the EEG signals. The MAHNOB-HCI-Tagging database was used for experimental purposes, which was collected over 27 participants with film clips as emotional stimuli. An averaged correct classification rate of 64.74% and 62.75% were achieved respectively on the 3-class Arousal and valence states classification problem using support vector machine (SVM) with ANOVA as feature selection mechanism. The second experiment, a proof-of-concept study, was to examine the suitability of the current in-market consumer-grade EEG headsets, with emphasis on the location of the electrodes, for the above affective states classification application.