A New Dataset and Recognition for Egocentric Microgesture Designed by Ergonomists.

Virtual and Augmented reality (VR/AR) are widely deployed in industrial, medical, educational, and entertaining fields. The design of interactive interfaces has an impact on usability, comfort, and efficiency. Hand controllers and gestures are popularly used in VR/AR devices. However, users may suffer from overloading on the upper extremities while raising the hand or controller. Therefore, we released a microgesture library with 19 microgestures designed by ergonomists. Users can perform microgestures for an extended duration by resting the forearm on the tables to reduce the load on the upper extremity. Additionally, we collected a microgesture dataset of 2900 samples and utilized the C3D model to recognize the microgesture dataset. Finally, we achieved a recognition accuracy of 93.4% on the microgestures dataset.