[4E3-OS-19b] OS-19 (2)

[4E3-OS-19b-04] Prediction of Listeners' Emotion Related to Speakers' Utterances during Discussion Using Multimodal Sensors

〇Motoki Sakai¹ (1. Tokyo Denki University)

Keywords:Group discussion, multimodal sensor

In higher education, there are many types of lectures, and increase of various abilities is demanded. A group discussion (GD) is one of activities conducted in university, and requires high communication skills of students. The goal of this research is to predict listeners’ emotions toward speakers’ utterances with electrocardiogram (ECG), electromyogram (EMG), and accelerometer. In this research, nine GDs were conducted with 20 students, listeners’ basic six emotions toward speakers’ utterances during GD were annotated. 56 features computed from sensor data were calculated. To investigate significant differences among features divided by six groups, the Kruskal-Wallis and multiple comparison tests were used. As a result, it was shown the possibility to discriminate between positive and negative emotions.