複数の非言語情報を利用した嘘の読み取りとその自動化

概要

論文の詳細を見る
Interactive and autonomous agents are considered to participate in our daily life in the future; in this situation, we expect that the agents have ability to communicate with people naturally. For natural communication, the agents should speculate about humans' intentions. For speculating intentions like deception, in this study, we focused on unconscious expressions when people tell a lie. Then, we experimentally investigated whether "lies" in a situation similar to actual communication were discernible by using nonverbal information such as gaze, prosody and facial expressions. We found that a discriminant analysis achieved 75-85% success rates in discriminating the participants' lies using 4 of 13 variables. The selected variables contain all the modalities to which we paid attention. The selected variables changed with situations and individuals. The discriminant function discriminated about 80% of an unknown data set of lies that was acquired for same people under a similar condition. For discrimination of lies by agents, it is neccessary to automate the measuring. For this purpose, we made an automatic measuring system of gaze directions and facial feature points. Then, we conducted an experiment by using the system. Therefore, we could get a clue that agents had a potential to speculate about human's intentions.
ヒューマンインタフェース学会,東京大学大学院総合文化研究科,Department of General System Studies, The University of Tokyo,NTT第二部門,NTT Department II,はこだて未来大学,Future University-Hakodateの論文
2006-11-25

複数の非言語情報を利用した嘘の読み取りとその自動化

スポンサーリンク

概要

著者

関連論文

スポンサーリンク