Emotion Recognition Using Electromyogram Recorded around the Eyes for Virtual Reality Applications

Abstract

Emotion plays important role in daily life, affecting social interactions, decision making and mental health. This makes emotion recognition technique to be highlighted as a crucial element of human-computer interface (HCI). Emotion recognition can be applied in various fields, and virtual reality (VR) is one of the upgrowing industry which can apply these applications. Real-time emotion recognition can be applied in VR environment such as video games, social network, education and neuromarketing. Although recent studies used biosignals such as electrodermal activity (EDA), heart rate variability (HRV), and electroencephalography (EEG), they have drawbacks of their own and facial electromyography (fEMG) can solve the problems. However, two major issues of fEMG emotion recognition exist: difficulty of wearing devices and long calibration time. In this paper, we proposed a new fEMG-based emotion recognition method for VR application by utilizing electrodes around the eyes and using no visual or auditory stimuli for calibration. We collected fEMG data around the eyes while participants voluntarily made facial expressions. Collected data were used for classification model training. For test data set, participants watched audio visual stimuli while fEMG data were recorded. Test data set was classified by the model and we evaluated the result. We found that linear discriminant analysis (LDA) classifier trained with dimension reduction using 27 s data showed 81.15% accuracy.