Emotion Recognition using Electroencephalography in Response to High Dynamic Range Videos

Abstract

Emotions are the human's internal feelings that result in physiological and physical changes influencing behaviour and thought. Emotion recognition is critical in affective computing for user modelling and human-computer interaction. Previously, user's emotional states have been recognized by measuring physiological activities in response to various audiovisual standard dynamic range (SDR) emotional stimuli. High dynamic range (HDR) content provides a better visual and immersive experience than SDR stimuli due to more brightness, enhanced contrast, and saturated colors. However, the impact of HDR multimedia on emotion elicitation and recognition has not been studied yet. In this paper, four audio-visual HDR stimuli were selected that evoke discrete emotions in each quadrant of the arousal-valence plane. Electroencephalography (EEG) signals of 27 subjects were recorded while watching HDR stimuli using a commercially available Emotiv Insight headset. Time-domain features from 5 channels of EEG signals are extracted and selected to classify emotions in two arousal and valence states using a support vector machine (SVM) classifier. Classification accuracies of 80.55% and 70.37% are achieved for arousal and valence, respectively. Our findings show that HDR videos can act as powerful stimuli for emotion recognition.