| The surface electromyography(sEMG)signal can be regarded as a new control source,which contains control information that can be used to control the upper limb rehabilitation robot system for designed robot functions.It was applied to the 5-DOF exoskeleton upper limb rehabilitation robot to recognize body’s intent.In this paper,the feature extraction methods of sEMG are studied,and the main contributions are as follows:First,we analyse the mechanism of sEMG and build a signal collection system based on the characteristic structure of sEMG.Considering the motion artifact noise in the acquisition process,the ButterWorth high pass filter is designed for denoising.Secondly this paper proposes a deep learning model for high dimension sEMG in feature extraction,extracting discriminant information effectively in a non-artificial manner.Multiple extreme learning machines autocoder(ELM-AE)reconstruct original sEMG in unsupervised training mode to achieve distributed expressions.The weights of the hidden layer represent the characteristics of sEMG,so that the compressed ELM-AE can map the high dimensional signal to the low dimension and filter the noise,which simplifies the classifier design.The effectiveness of the method is proved to be better than the time and frequency-domain features in the simulations.Finally,in order to get visual feedback in the course of rehabilitation training,the virtual reality environment is built for the exoskeleton upper limb rehabilitation robot system in this paper.According to its overall structure and kinematics model,the virtual upper limb rehabilitation robot is drawn by the 3D MAX software.The upper limb rehabilitation robot model is invoked in the virtual reality toolbox of MATLAB to carry out visual simulation,based on which an interactive virtual reality game is designed.This environment can interact with patients,helping to interest patients to rehabilitation training,and improve the safety of the system. |
PDF Full Text Request