The Design Of Electromyography Training And Testing Platform Based On Virtual Environment

Myoelectric control prosthetics,combining both pattern recognition and multi-DOF mechanical hand,helps amputees restore arm/hand functionality and aims at improving their life quality.Amputee's residual muscles,however,cannot provide EMG signals as well as normally-limbed people,and thus they usually have lower classification rate and do not want to use myoelectric prosthetics.For this reason,many training solutions have been proposed,among which training within virtual environment is one of the most popular researches.In this thesis,representative works have been discussed at first.Based on common features and shortcomings,the thesis proposed that the virtual training environment should have visual feedback,tactile feedback and high extendibility,and an architecture implementing these concepts is presented.According to topological structure of human hand and prosthetics,a general modeling strategy for tree-like prosthetics has been proposed,and its abstract interface has been designed.According to two major forms of motion control,discrete actions and continuous joint angles,abstract control interfaces of control characters and extern joint data have been designed.Not only has the software built 4 virtual prostheses and 3 control strategies,but it also allows further extension without changing platform code.These interfaces allow importing arbitrary prosthetics with tree-like topology.A pattern recognition sub-system has been implemented in this thesis,including a training module and a testing module.The training module can be used to create,import and export action series,and to collect data.And testing module,using a "plug-and-play" principle,can connect to any algorithm whose extern interfaces conform to requirements.Through experiments,using LDA and SVM plugins,extensibility in algorithms has been proved.To meet the requirement of interactive grasping,kinematic response between virtual objects has been implemented using contact points and normal vectors.In order to make virtual environment more realistic,dynamic response between virtual objects has been implemented using a physics engine.And a virtual sensor has been designed to measure contact force between objects,so that the contact force information can be fed to users.Experimental results have validated the effectiveness of tactile feedback.