Human Pose Estimation from Egocentric Videos

According to a survey conducted by the Ministry of Health, Labour and Welfare in 2019, about 30% of patients suffer from back pain and stiff shoulders.Although researches on pose estimation have been conducted for a long time, they cannot be used for daily pose estimation, because they need fixed cameras to capture target/subject motion.To solve this problem, the present paper, proposes a novel pose estimation method from egocentric videos using Epipolar Geometry.It computes three rotational angles, i.e., pitch, yaw and roll, from the egocentric motion videos to evaluate differences from his/her normal motion.In the experiment, three egocentric videos were used to verify the performance and effectiveness of the proposed method and reasonable/satisfactory results were obtained.