Features of Event-related Potentials Used to Recognize Clusters of Facial Expressions

To assess human emotion using electroencephalograms (EEGs), the relationship between emotional impressions of images of facial expressions and features of Event Related Potentials (ERPs) recorded using three electrodes was analyzed. First, two clusters of emotional impressions were extracted using two-dimensional responses of the Affect Grid scale. Second, features of ERPs in response to the two clusters were examined. Time slots where amplitude differences in ERP appeared were measured, and differences in the frequency power of ERP were also extracted for each electrode. To evaluate these features, prediction performance for the two clusters was examined using discriminant analysis of the features. Also, the dependency of some band pass filters was measured.