Overlap Makes Perfect: Designing a Meta-Tutor With Live Corrective Feedback for Ubiquitous Motion Learning

It has been deemed as a fundamental difficulty in the task of motion learning for MOOC courses, due to the limitation of MOOC’s defining feature of online teaching. This paper proposes a novel method called Meta-tutor for MOOC motion learning, by using neural style transfer and AR techniques to overcome the above problem. The Meta-tutor constructed by neural style transfer discards the original identity of real-world tutor in the MOOC video and can be regarded as a Meta-human in the Metaverse of MOOC’s virtual learning space. A light-weight wireless framework is designed to support ubiquitous motion learning, where the learner can present at any real-world practising field (e.g., sports court, dance room and training yard) and only needs take a smart mobile phone and wear a commodity AR device. This research is the first step towards a long-term study program.