Research On Configuration Optimization And Virtual Training System Of The Sensory-integrated Operational Prosthesis Hand

In order to help new prosthesis hand users take EMG control training for graspmotion, assist and guide the new prosthesis hand with its design, debugging,performance verification and product optimization, funded by national “973” project“Design Manufacturing and Performance Evaluation of the Sensory-integratedOperational Dexterous Prosthesis(No.2011CB013306)”, an optimized design methodfor the configuration of prosthesis hand and analyses towards the grasp state of virtualprosthesis hand based on form closure theory was explored. Meanwhile a completehigh-immersion virtual training system for the sensory-integrated operational prosthesishand was designed, which integrated with EMG collection, EMG decoding for graspintention, grasp controlling, collision detection and virtual torque-position sensor, andperformance evaluation of the virtual prosthesis hand from EMG decoding time, EMGdecoding success rate and motion comparison between prosthesis hand and human handwas carried out.Firstly, a configuration optimized design method for the five finger hand wasproposed based on the form closure theory, considering the multi-point contact graspcharacteristics of the human hand. Through a J discriminant function the form closurestatus of power grasps with multi-parameter input for the five finger hand wascalculated, after determining the necessary parameters of the configuration optimizeddesign for the five finger hand according to the grasp movement requirements. Then aninterval iterative algorithm was used to achieve the feasible interval maximization of theparameters for the five fingers on the premise of form closure, with norm domain widthbeing its parameters optimization performance indicator. The best design results of themost important5parameters of the five fingers hand was obtained from a sphericalGrasp as an example. Simulation results showed that performance indicator afteroptimization was16%higher than the previous one. Not only the optimized results canguide configuration design of the prosthesis hand, but also determine the grasp state ofthe virtual prosthesis hand.Secondly, based on VC++/Open Inventor, the virtual training system for prosthesishand was built, which has integrated operation and aesthesia for the virtual prosthesishand. Operation of the virtual prosthesis hand consists of EMG collection through NI-DAQ, interface with VC++, EMG decoding algorithm for grasp intention according toreal prosthesis hand, and grasps controlling based on universal finger dynamics. Whileaesthesia of the virtual prosthesis hand consists of collision detection between virtualprosthesis hand and virtual objects, determination rules of success grasp based oncollision and form closure theory, and virtual torque-position sensor based on universalfinger statics of the virtual prosthesis hand and experiment results on real torque sensor of the prosthesis hand.Finally, performance evaluation of the virtual training system for prosthesis handwas accomplished, from EMG decoding time, EMG decoding success rate, and keykinematic performance comparison between virtual prosthesis hand and human hand.Contents and calculation algorithm of the EMG decoding time were analyzed, andexperiment on EMG decoding for virtual prosthesis hand grasp motion was designedand carried out, whose results turned out that time range and success rate of EMGdecoding in6basic grasps were the same with real prosthesis hand. Experiment onhuman hand grasps based on VICON was designed and carried out,2kinematic indexesbased on joint angle information and fingertip trail during the grasp process werecreated to access the difference between virtual prosthesis hand and human hand in bothpower grasp and precision grasp.