Perception of Emotional Expression of Mobile Humanoid Robot Using Gait-Induced Upper Body Motion
Perception of Emotional Expression of Mobile Humanoid Robot Using Gait-Induced Upper Body Motion
复制标题
利用步态诱导的上半身运动感知移动人形机器人的情感表达
DOI：
10.1109/access.2021.3110160
复制
发表时间：
2021
期刊：
影响因子：
3.9
通讯作者：
Ishiguro Hiroshi
中科院分区：
文献类型：
--
作者：
Yagi Satoshi;Nakata Yoshihiro;Nakamura Yutaka;Ishiguro Hiroshi
人类不仅可以通过面部表情或手势表达自己的情绪，还可以通过全身运动来表达自己的情绪，甚至在走路时也是如此，这对我们的人际关系和与他人的互动至关重要。与人交流也被认为是机器人的一项重要任务，但轮式移动的人形机器人在移动时的情感表达尚未得到很好的研究。在本文中，我们在一个移动的android ibuki中实现了情感类人步态诱导的上半身运动。我们假设，步态引起的上身运动与垂直振荡增强了人类感知的移动的机器人的情感表达。Ibuki是一种具有垂直摆动机构的移动的机器人，该垂直摆动机构使得机器人能够利用人类步态引起的上身运动来移动其身体。我们进行了两个实验，参与者观看了ibuki在一个房间里移动的视频，表达了不同的情绪-愤怒，快乐和悲伤-然后回答他们如何感知这种情绪以及他们各自的信心水平。我们的研究结果表明，对于垂直振荡的运动，识别率为：56.0%的愤怒，77.7%的快乐，和97.0%的悲伤。我们还发现，表示幸福的垂直振荡的动作的识别率和置信水平较高。我们的研究结果在为人类和机器人之间更自然的互动方式开辟道路方面发挥了重要作用。例如，通过在表达情感时更接近人类，机器人自然可以提供与他们开始对话的机会。
Humans can express their emotions with not only facial expressions or gestures but also whole-body motion, even while walking, which are essential for our interpersonal relationships and our interaction with others. Communication with people is also considered an important task for robots, however emotional expressions by wheeled mobile humanoid robots when moving has not been well studied. In this paper, we implemented emotional human-like gait-induced upper body motion in a mobile android ibuki. We hypothesize that the gait-induced upper body motion with vertical oscillation enhances the human perception of a mobile robot's emotional expressions. ibuki is a mobile robot, which has a vertical oscillation mechanism, which enables the robot to move its body with a human gait-induced upper body motion. We ran two experiments in which participants watched videos of ibuki moving in a room expressing different emotions - anger, happiness, and sadness-, before answering how they perceived the emotion and their respective confidence level in their answer. Our results show that for motions with vertical oscillation, recognition rates were: 56.0 % for anger, 77.7 % for happiness, and 97.0 % for sadness. We also found that the recognition rate and confidence level of motions expressing happiness with vertical oscillation were higher. Our results play an important role in opening up avenues for more natural ways of interaction between humans and robots. For example, by approaching humans closer while expressing emotions, a robot can naturally provide an opportunity to get started conversation with them.