Hand Gesture Recognition in Multi-space of 2D/3D

Recent human-computer interaction studies focus on natural interface based on hand/finger movements by tracking body motion using various sensors. In this paper, we propose 'Surface Gesture' recognition method using both 2D surface information and 3D spatial information. 'Surface Gesture' is composed of continuous gestures using 2D and 3D coordinates in multi spaces. We use a touchscreen to obtain the 2D coordinate and depth sensors to obtain finger coordinates in 3D space. We decompose 'Surface Gesture' into 3 particular steps and recognize 'Surface Gesture' with tagging a label on each step. By combining accurate 2D plane information and natural 3D space gestures, our method provides a more convenient and familiar way to interact with computers. We conducted an experiment to evaluate performance of the recognition engine. Experimental results show that the proposed method provides a 94% recognition rate on particular step and an 82% recognition rate on 'Surface Gesture'