User Interfaces Visualization design and evaluation methods
Beschreibung:
In this paper, we present our first prototype of a virtual group meeting emotion analytics tool based on both real human participants and virtual agents. The analysis starts with the detection of facial expressions of every participant using Facial Action Units (FACS), and calculates an aggregated summary of the group’s emotional states throughout the duration of the meeting. We infer the intensities of the following primary emotional states: happy, surprised, angry, sad, disgusted, fearful by using the known AU combinations. Additionally, an AI algorithm was trained and tested on the DAiSEE dataset to infer the following secondary emotional states: engaged, bored, confused and frustrated. We present this tool in the context of an online classroom scenario in which the teacher can receive an anonymized aggregated emotion analytics of his/ her student group after their lecture calculated for the entire duration of the lecture. We finally asked 21 participants to rate the usability of this tool.