Confidence Identification Based on the Combination of Verbal and Non-Verbal factors in Human Robot Interaction

• Main Authors: Hsieh, Wei-Fen, Li, Youdi, Kasano, Erina, Simokawara, Eri-Sato, Yamaguchi, Toru
• Format: Conference Proceeding
• Language: English
• Subjects: computational human-robot interaction; Computer science; confidence detection; Deep learning; human-centered computing; Human-robot interaction; Robot kinematics; Robot sensing systems; robotics; System analysis and design
• ISSN: 2161-4407
• DOI: 10.1109/IJCNN.2019.8851845

Not only verbal information but also Non-verbal information is an essential factor in Human-Robot Interaction (HRI). In order to understand Human partner's perception, confidence status detection plays an important role for the robot to be capable to provide further information in the conversation. Moreover, the confident/uncertain feature can be utilized on robot expression to smooth the interaction and gain social skill. On another aspect, certain/uncertain expression style makes it possible to form extraversion/introversion personality which might spice up the interaction. This paper presented the concept of confident expression in HRI and an experiment to analyze both verbal and non-verbal features for confident expression. The confidence conditions were defined as 5-level status according to the questionnaire answer: most confident, comparative certainty, neutral, relative uncertain, most unconfident. The results showed that verbal and nonverbal features are possible to be classified into most confident/uncertain condition with the accuracy of 73.2143 % analyzed by multinomial logistic regression model via Weka. However, the ambiguous differences between the expression style are difficult to be classified. The accuracy of predicting comparative certainty and relative uncertain status can merely reach 66.67%. The classification of comparative certainty and relative uncertainty still need more features to increase prediction accuracy.