Finger Multi-Joint Trajectory Measurement and Kinematics Analysis Based on Machine Vision

Abstract

A method for measuring multi-joint finger trajectories is proposed using MediaPipe. In this method, a high-speed camera is used to record finger movements. Subsequently, the recorded finger movement data are input into MediaPipe, where the system automatically extracts the coordinate data of the key points in the finger movements. From this, we obtain data pertaining to the trajectory of the finger movements. In order to verify the accuracy and effectiveness of this experimental method, we compared it with the DH method and the Artificial keypoint alignment method in terms of metrics such as MAPE (Mean Absolute Percentage Error), maximum distance error, and the time taken to process 500 images. The results demonstrated that our method can detect multiple finger joints in a natural, efficient, and accurate manner. Then, we measured posture for three selected hand movements. We determined the position coordinates of the joints and calculated the angular acceleration of the joint rotation. We observed that the angular acceleration can fluctuate significantly over a very short period of time (less than 100 ms), in some cases increasing to more than ten times the initial acceleration. This finding underscores the complexity of finger joint movements. This study can provide support and reference for the design of finger rehabilitation robots and dexterous hands.