The goal of Dongsheng Yang's work is to find the optimal Robot Facial Expressions (RFEs)
automatically and objectively. He used a facial android robot (named Nikola) with 35 pneumatic actuators, which means the robot is very complex with 35 degrees of freedom. To deal with this expensive black box problem, he used the Bayesian Optimization algorithm with a computer vision-based emotion classification model and generated a set of optimized robot facial expressions. After that, he also did the human evaluation experiment, which showed the improved expressions have better performance in Anger, Disgust, Sadness, and Surprise.
Robot facial emotion design, human-robot Interaction
| 氏名 | 専攻 | 研究室 | 役職/学年 |
|---|---|---|---|
| Dongsheng Yang | 知能情報学専攻 | 認知情報学 | 博士1回生 |
| 佐藤 弥 | 知能情報学専攻 | 計算論的認知神経科学連携ユニット | 教授 |
| Qianying Liu | 知能情報学専攻 | 言語メディア | 博士3回生 |
| 港 隆史 | その他: その他 | RIKEN | その他: その他 |
| 難波 修史 | その他: その他 | RIKEN | その他: その他 |
| 西田 眞也 | 知能情報学専攻 | 認知情報学 | 教授 |