Detecting Human Attention States Exploiting an Intelligent Approach Combining Fuzzy Inference System and Support Vector Regression

In recent years, one of the main issues in e-learning has been the personalization of educational content for learners. Identifying cognitive factors that influence learning plays a significant role in improving the performance of personalization systems. “Attention level” is one of these factors. Given the absence of a teaching agent in the e-learning environment, detecting learner attention in this context is very complex. Therefore, designing intelligent systems that can detect learner attention and provide appropriate educational feedback is very important. This research estimates individuals’ attention levels by analyzing “head movements.” To this end, the head pose angles, including Pitch, Yaw, and Roll, were estimated using the Support Vector Regression (SVR) algorithm and optimized through a Particle Swarm Optimization (PSO) algorithm. Following these three metrics considered input variables for the fuzzy inference system, a system was designed to estimate attention levels. Applying this system to the AFLW2000 dataset yielded an accuracy of 86%, which significantly outperformed previous studies.