Real-time Motion Synchronization between Different Persons using Similarities of 3D Poses

Summary

In this research, we focus on the technology that plays two similar motions by different persons in synchronization. The proposed method finds timing from the recorded motion data, which is matched with the inputted user's pose, in real-time. Real-time motion synchronization allows a user to obtain a pose instantaneously which is closest to the user's present pose during making motions. As an promising example application, we propose Synchronized Video.

For the synchronization between different motions, the proposed method uses 3D information of the poses. By using 3D information, the synchronization is expected to be achieved without relying on camera specification and setup configuration. On the other hand, we need to solve the problem caused by the difference in stature and skill level. Therefore, the method introduces new recognition features and matching technique which is not affected by such differences in motions.

Similarity map of motions and the synchronized poses

Movie