假手交互控制系统及基于压力分布的多动作模式识别分析-artificial hand interactive control system and multi-action pattern recognition analysis based on pressure distribution.docx

的压力分布特征。比较实验和可分性分析表明， PCA 法和 DQPSO 法在获得了明显的降维效果的同时，保证了良好的可分性。 进一步的，为了实现对所获得的多种手部动作模式特征准确可靠的识别，本文提出基于 KNN 算法、SVM 算法和 BPNN 算法设计特征分类器，用于识别 手部动作的压力分布特征。经交叉验证实验比较表明，三种分类器中，KNN 分 类器具有最好的识别成功率和稳定性。假手的在线识别控制实验显示，使用 KNN 分类器可以实现对人手 17 种动作准确快速的识别。关键词：假手；交互控制系统；压力分布；模式识别；电刺激；感觉反馈AbstractIn order to mimic the human hand in terms of shape, structure, flexibility, a lot of prosthetic hand with humanized shape, size and multi joints/degrees of freedom have been developed in the recent years,. On the other hand, the research on biderictional interaction between the prosthetic hand and the human body is insufficient. There is no effective method on accurately identifying the various modes of human hand movements, and reliablely feeding back the sensory from the prosthetic hand to the human body. Focus on the interactive control for the multi-DoFs prosthetic hand, this paper presents an interactive control system between the prosthetic hand and the human body. And based on this control system, a new method for the multi-motion recongnition is proposed. The contents of this paper includes: the development of high-performance, integrated prosthetic hand control system, development of the human-machine interface based on the pressure distribution and the electrical stimulation, investigation on feature extraction methods and identification methods based on the the pressure distribution features of hand movements, and experimental researches.First, overview about newly developed multi-DoFs prosthetic hands and typical interactive control system and control strategies are given in details. Then, a new interactive control system structure of the prosthetic hand is designed, which is divided into two parts: the prosthetic hand control system and the human-machine interface. In detail, the prosthetic hand control system consists of sensor system and motion control system. The sensor system is composed of the joint torque sensors, joint position sensors and motor encoders. These sensors provid