FST Graduate Research Topic

Research Topic: Students' engagement detection in online learning
Suggested By: Ali Dewan
Last Update: April 17th, 2025

In distance education, the learning environments (e.g., Learning Management Systems (LMS)) aim to deliver efficacious instruction, but rarely take into consideration the motivational factors involved in the learning process. However, the motivational factors like engagement play an important role in effective learning. Learning environments could be improved by tracking students’ engagement that, in turn, would allow personalized interventions. This idea has been exploited by some Intelligent Tutoring Systems (ITS) in some extent, but not yet in any learning environments (e.g., Moodle) that are typically used in distance education. To address this gap, this research will look at computer vision and artificial intelligence techniques for students’ engagement detection in online learning-content-delivery systems. In some previous research, several attributes relevant to engagement detection were identified by means of log-file analysis on web-based learning environments. In this research project, we will explore how computer vision and artificial intelligence techniques can be applied to detect students’ engagement while students involve in a learning activity in distance education settings, such as reading, writing, participating in online meetings, and watching online video tutorials like the activities typically encountered in campus based regular educational settings. Students’ engagement will be annotated in different levels ranging from completely not engaged to strongly engaged. Computer vision techniques – Gabor, Histogram of Oriented Gradient (HOG), and Local Binary Pattern (LBP) – will be used to extract features from the videos and feed into three different classifiers – Support Vector Machines, Bayesian Network, and Convolution Neural Network (a deep learning-based approach) for the engagement detection. Area under the Receiver Operational Characteristics (AU-ROC) and two-alternative forced choice (2AFC) will be used to evaluate the system accuracy using some popular benchmarking datasets, FER2013, CMU Multi-PIE, and CK+. AU Moodle courses and AlgoVis (an algorithm visualization tool developed by Prof. Oscar Lin and his team) will also be used as the testbed for the students’ engagement detection.  This research topic is suitable for thesis, project, and essay route students.