An Impression Evaluation of Robot Facial Expressions Considering Individual Differences by Using Biological Information

The application of the robot services has been in high demand due to the declining birthrate and aging society. However, one robot's expression could have different perceptions. Therefore, the purpose of this study is to improve the robot's impression by estimating human emotions using biological information. Here, a machine learning method was proposed to consider individual differences from the impression evaluation and the combined measurements of electroencephalography (EEG) and pulse rate. This method was evaluated based on three patterns of robot expressions: same as human emotion (synchronize), against human emotion (reverse synchronize), and funny facial expressions. A machine learning method was implied to create a classification model to decide the facial expression of a robot that suits users' preference. As a result, the individual differences were observed and the machine learning approach reached 80% accuracy.