3D Articulated Human Motion Analysis System Using a Single Low-Cost RGB-D Sensor
Jong Sung Kim, Myung-Gyu Kim · 2018
In this paper, an cost-effective, highly accurate 3D articulated human motion analysis system is proposed. In the proposed system, a single low-cost RGB-D sensor captures a color image and depth one of human motion. Then, a deep learning-based 2D motion analysis process accurately estimates intermediate 2D articulated human motion from the color image. Finally, a color-to-depth warping-based 3D motion transform process effectively compute final 3D articulated human motion from the depth image. The proposed system is cost-effective but highly accurate. Experimental results show that the proposed system outperforms the commercial system using the same RGB-D sensor when compared in terms of accuracy.