Posture Classification and Writing Behavior Estimation in Individual Learning Using OpenPose
Daisuke Kushida, Raito Suzuki · 2023
The development of remote learning environments is being promoted to provide equal opportunities for learning. However, some challenges exist with remote learning, such as difficulty in learners maintaining their concentration and the inability to determine how a learner state of being. In this study, the authors propose a new method to monitor learners’efforts in a remote class while avoiding communication loads, such as video images. The authors use Open- Pose detection library to obtain the body skeletal coordinates of an individual learner’s frontal Red-Green-Blue color (RGB) model video, including the learner’s hands, and attempted to classify the learner’s posture by pattern matching, and estimate the learner’s writing behavior using wrist coordinates. Consequently, both posture classification and writing behavior estimation were possible with an accuracy of more than 80%, suggesting the possibility of perceiving a remote individual learner’s poise without using video communication.