The Study of Traditional Chinese Medicine Manipulation Simulation Based on Gesture Recognition of a Dexterous Five-Finger Robotic Hand

This study utilizes human-computer interaction to simulate the traditional Chinese medicine techniques with the dexterous hand. First, a high-precision gesture recognition system based on Mediapipe and support vector machine (SVM) is constructed. Mediapipe is responsible for identifying the 3D coordinates of the 21 joints in the hand, and the standardized data is processed through the calculation of joint angle and distance features to extract hand features. Then, the radial basis function (RBF) kernel SVM is used for feature classification, and the one-to-one strategy is used to extend the binary SVM to a multi-class model, thus achieving the classification and recognition of traditional Chinese medicine techniques. The experimental results show that the system can simulate the traditional Chinese medicine techniques of the human hand very well.