An Emotion Recognition System Based on Physiological Signals Obtained by Wearable Sensors

Automatic emotion recognition is a major topic in the area of human--robot interaction. This paper presents an emotion recognition system based on physiological signals. Emotion induction experiments which induced joy, sadness, anger, and pleasure were conducted on 11 subjects. The subjects’ electrocardiogram (ECG) and respiration (RSP) signals were recorded simultaneously by a physiological monitoring device based on wearable sensors. Compared to the non-wearable physiological monitoring devices often used in other emotion recognition systems, the wearable physiological monitoring device does not restrict the subjects’ movement. From the acquired physiological signals, one hundred and forty-five signal features were extracted. A feature selection method based on genetic algorithm was developed to minimize errors resulting from useless signal features as well as reduce computation complexity. To recognize emotions from the selected physiological signal features, a support vector machine (SVM) method was applied, which achieved a recognition accuracy of 81.82, 63.64, 54.55, and 30.00 % for joy, sadness, anger, and pleasure, respectively. The results showed that it is feasible to recognize emotions from physiological signals.