Kinect-based dynamic head pose recognition in online courses
Zijian Fan, Jing Xu, Wei Liu, Feifan Liu, Wenqing Cheng · 2016
With the development of e-Learning applications, there is an increasing demand to measure or evaluate learners' behavior in online courses. Various technical approaches have been proposed for this purpose, among which the measurement of the learner's head pose is a fundamental technique. It can be used to evaluate the learner's attention, e.g. whether he/she is reading the online course materials. Considering the learner may move his position or change his gesture, there is a technical requirement to estimate the learner's facing angle to the monitor screen, despite the learner's movement. In this paper, we propose a dynamic head pose recognition method, based on the RGB-D data measured by the Microsoft Kinect device. After obtaining the position of learner's head, this method estimates the angles of learner's head, dynamically calculate the proper angle threshold standing for focusing on the monitor screen, and finally output the recognition of learner's attention state. We conducted numerous experiments, and found that the proposed method could recognize the studying status from the learner's head pose during his movement.