접근센서 기반 물체 점구름 정보를 활용한 의수 사용자 의도 인식 및 자율제어 시스템

Abstract

Functional prosthetic hands have been developed to recover the functionality of hand amputee patients. Various mechanisms have been applied to implement the hand motion. To control the motion of the prosthetic hand, the users need the interface to transfer their intention. Surface electromyography(sEMG) is widely used, but it has limitations to represent all dexterous movements of the hand. In this research, we demonstrate the novel method that uses the relative relationship between the object and the prosthetic hand, to infer the proper grasp posture. The developed system consists of the proximity sensor array in the palmar side of the prosthetic hand and the posture tracking sensor (T265). By combining the data from those sensors, the system generates the point cloud of the target object during the approaching motion of the prosthetic hand. Based on the point cloud data, the system determines the target grasp posture intended by the user and the size of the object. After the post-processing that calculates the optimal contact points, the system gives the motion command for the grasping. All the processes are done in real-time and have less than 100ms delay. It shows 96% task classification accuracy between power, precision pinch, and lateral pinch grasp.