複数台RGB-Dセンサを用いた片麻痺者の状態推定システム

―複数台RGB-Dセンサ情報の統合―

抄録

To judge and suggest optimal training for realizing early rehabilitation, the estimation using multiple RGB-D sensors is proposed. The system consists with multiple RGB-D sensors and measuring computers connects with each RGB-D sensor, a server, which integrates information from each computer. Based on measurements of a target person by a RGB-D sensor, each computer estimates positions of all joints of the target. The server collects and integrates joint information from each computer. By integrating information, the server compensates occlusion. Based on the integrated joint information, the system judges the grade of paralyze and the state of the target and suggests training to the target, family and so on, supporting menu to an other system or supporting robot/robots. In this paper, a method for position calibration of each RGB-D sensors and a method for integrating information measured by each sensor are described especially on Microsoft KINECT V2. By using implemented experimental system, the validity of the proposed method is shown.