Innovative Teaching Method for Dunhuang Dance Incorporating Pose Estimation Fusion

Abstract

To overcome the challenges of high costs, venue requirements, and lack of portability in current digital teaching for Dunhuang dance, this study uses Vicon optical motion capture to create a Dunhuang dance motion database and virtual dancers. Using lightweight Convolutional Neural Networks (CNN) for single-view pose estimation in Unity, it captures students' 3D body movements in real-time, driving their virtual avatars. The result is a virtual scene replicating the "one-on-one" teacher-student interaction. This approach enhances user engagement, offering a fresh learning experience and introducing a new teaching paradigm for Dunhuang dance.