No abstract

No abstract

No abstract

In this paper, an EEG-based emotion database was reconstructed using spectral subtraction, and recognition performances were evaluated. For subtraction, we created two types of databases. One database included facial expression readings, and the other included both emotion and facial expression readings. A reconstructed database containing pure emotional information was achieved by spectral subtraction, and compared with the original recorded data of emotion and facial expression readings. Facial expression illustrations and the International Affective Picture System were used for inducing facial expressions and feelings. EEG data was recorded after an emotion was excited or while imitating a particular facial expression. By subtracting the database of information related to facial expressions from a database about facial expressions and emotions, pure information about emotion was created. The method used to separate emotion and expression in a database was spectral subtraction. Recognition experiments were classified into six types of emotions. Using the original database, the true emotion could be guessed from EEG readings 29.9% of the time, but using the reconstructed database resulted in an 81.7% recognition rate.