Abstract
E-learning is becoming increasingly more popular. However, for such platforms (where the students and tutors are geographically separated), it is necessary to estimate the degree of students' engagement with the course contents. Such feedback is highly important and useful for assessing the teaching quality and adjusting the teaching delivery in large-scale online learning platforms. When the number of attendees is large, it is essential to obtain overall engagement feedback, but it is also challenging to do so because of the high levels of uncertainty associated with the environments and students. To handle such uncertainties, we present a type-2 fuzzy logic based system using visual RGB-D features including head pose direction and facial expressions captured from a low-cost but robust 3D camera (Kinect v2) to estimate the engagement degree of the students for both remote and on-site education. This system enriches another self-learning type-2 fuzzy logic system which provides the instructors with suggestions to vary their teaching means to suit the level of course students and improve the course instruction and delivery. This proposed dynamic e-learning environment involves on-site students, distance students, and a teacher who delivers the lecture to all attending onsite and remote students. The rules are learned from the students' behavior and the system is continuously updated to give the teacher the ability to adapt the lecture delivery instructional approach to varied learners' engagement levels. The efficiency of the proposed system has been evaluated through various real-world experiments in the University of Essex iClassroom on a sample of thirty students and six teachers. These experiments demonstrate the efficiency of the proposed interval type-2 fuzzy logic based system to handle the faced uncertainties and produce superior improved average learners' engagements when compared to type-1 fuzzy systems and nonadaptive systems.