Remarks on computational emotion recognition from vital information

This paper investigates computational emotion recognition using multi-modal bio-potential signals. Two vital signs, pulse and skin conductance response, are measured to evaluate three emotions: positive emotion(relax and pleasure), negative emotion(stressful and unpleasure), and normal. Psychological experiments using audio contents in order to excite emotions to subjects are undertaken to gather the bio-potential signals. For computational emotion recognition, support vector machine is used to design the emotion recognition system. In experiments of recognizing emotions, a recognition rate of 41.2% is achieved for three emotions by using the features including dynamics information extracted from the bio-potential signals. The results obtained in this study demonstrate that using multimodal bio-potential signals is feasible for computational emotion recognition and that SVM is suited for this task.