Estimation of Forearm Pose Based on Upper Arm Deformation Using a Deep Neural Network
Sung-Gwi Cho, Tetsuya Kurasumi, Masahiro Yoshikawa, Ming Ding, Jun Takamatsu, Tsukasa Ogasawara · 2019
Activities of body tissues such as muscles, tendons, and bones appear in the deformation of skins. It would be possible to obtain the information of the body motions by measuring the skin deformation. In our research, we focused on upper arm deformation, which is changed with not only the elbow flexion and extension but also the forearm pronation and supination. Thus, with the measurement of the upper arm deformation, we can estimate the elbow pose. In this paper, we propose a forearm pose estimation method based on upper arm deformation. A distance sensor array with eight sensor units is developed to measure the upper arm deformation. From the measured deformation, a deep neural network (DNN) based model is trained and used to estimate the forearm pose. The results of experiment show that our proposed method was able to estimate the forearm pose with RMSE of 2.9 degrees for the elbow joint angle and 7.6 degrees for the forearm pronation and supination angle.