Real-Time EEG-based Human Emotion Recognition

Abstract

Recognition of user felt emotion is an exciting field because visual, verbal and facial communications can be falsified more easily than ‘inner’ emotions. Non-invasive EEG-based human emotion recognition entails the classification of discrete emotions using EEG data. These emotions can be defined by the arousal-valence dimensions. We performed real-time emotion classification for four categories of emotional states, namely: pleasant, sad, happy and frustrated. Higuchi’s Fractal Dimension was applied on EEG data and used as a feature extraction method and Support Vector Machine was used for classification. This paper documents a comparative study of classification accuracy achieved by collecting raw EEG data from 3 electrode locations vs. collection from 8 electrode locations.