The Analysis of Nonverbal Behavior for Detecting Awkward Situation in Communication

Abstract

In recent years, robots that communicate with humans have been deployed in human's daily life. In the application, communication robot is expected to provide information or recommendation of products in market place. In the field of human robot interactions, previous research focuses on human's emotion recognition to improve robot's communication capability. Especially, detecting awkward situations during communication is one of the challenging tasks in human robot communication. In this paper, we propose a multimodal system to detect an awkward situation during communication. Our research focuses on nonverbal behavior information that expressed unconsciously during communication. We designed the experiment which makes participants feel awkward during the communication. Nonverbal behavior data that recorded during the experiment were analyzed to verify the awkward situation. The results showed that the detection accuracy of our proposed method could achieve about 80%.