Human Motion Assessment on Mobile Devices

Hobeom Jeon, Dohyung Kim, Jaehong Kim

Abstract: Human motion analysis is being performed in various places using various devices. However, the complexity of the analysis algorithms required by high-performance devices makes it difficult to run these algorithms anytime and anywhere. We have successfully ported a pose estimation model to mobile devices and applied various post-processing algorithms. Our analysis algorithm, which uses the joint coordinates of the pose estimation model, has both fast calculation speed and versatility. Our lightweight pose models and motion analysis algorithms allow users to analyze fitness actions and receive instant feedback in real time using their smartphones. Our motion assessment system is applied to real mobile devices and tested in the real world. Based on empirical research on real environments, our motion assessment system is shown to be applicable to various fields.