A Preliminary Experiment on the Estimation of Emotion Using Facial Expression and Biological Signals

Abstract

Imagine the day that a robot would comfort you when you feel sad. In the field of artificial intelligence and robot engineering, there are many research regarding automatic classification of human emotion to enhance human-robot communication, especially for therapy. Generally, estimating emotions of people is based on information such as facial expression, eye-gazing direction, and behaviors that are expressed externally and the robot can observe through a camera and so on. However, there is some invisible information that cannot be expressed, or control not to express. In this case, it is difficult to estimate the emotion even if the analysis technologies are sophisticated. The main idea of this research is to compare the classified emotion based on two different sources: controllable and uncontrollable expression. The preliminary experiments show that our proposed method suggested that the classification of emotion from biological signals outperform the classification from facial expression.