Deep Learning-Aided Head-Posture Monitoring via a Wearable Device for Learning Attention Application

Abstract

Learning attention determines educational outcomes and the effectiveness of online instruction. Conventional instructor-led interventions are inherently limited in digital environments, and vision-based head-posture monitoring suffers from poor robustness under real-world conditions. Here, we present a deep-learning-based framework that uses a wearable inertial sensor to continuously monitor head postures for learning attention application. A wearable six-axis device captures head kinematics. The model achieved an accuracy of 98.09% under the reported evaluation protocol. Leave-one-subject-out cross-validation was additionally used to assess performance on unseen participants. An attention-rating model maps these classes onto three attentional states based on trained postures. Our approach is unaffected by lighting changes and occlusion, enabling deep-learning-aided head-posture monitoring for learning attention application. This work establishes a methodological foundation for integrating wearable sensing and deep learning in education, with potential for personalized instruction.