Research On Intention Recognition And Gait Planning Of Exoskeleton Robot By EMG Signal

With the aggravation of the population aging problem in China,the number of elderly people with inconvenient lower limbs is increasing due to various diseases or weakened human functions.At the same time,there are still a large number of disabled people.Most of them have difficulty in action and cannot take care of themselves.There need a large number of medical personnel to work.In order to solve the shortage of professionals and labor costs,government began to attach importance to relying on advanced service robot technology to protect and improve people's health.Lower limb exoskeleton robots can significantly guarantee the movement ability and quality of life of people with lower limb dysfunction.Researching exoskeleton robots for helping the elderly and the disabled has important social and economic value for reducing family and social burden and promoting social harmony.At present,exoskeleton robots technology is still in the research and development stage in our country,and there is still no mature commercial application.The coupling performance between exoskeleton robot and human is still not perfect.In order to solve this problem,it is necessary to research the human-computer cognitive interaction systems.Intention recognition is a research hotspot of human-computer cognitive interaction.Given the physical interaction between our existing exoskeleton robots and humans,the key operations of exoskeleton robotic are complex and poorly sustainable.This thesis is based on a bioelectrical signal: the electromyographic(EMG)signal.And carries out research on the control mode of the EMG signal intention recognition combines with the finite state machine.EMG signal intention recognition and gait planning of exoskeleton robot are two main research points in this thesis.This thesis designs a strategy based on arm movement to control lower limb exoskeleton movement.This experiment collects EMG data of different movements of human arms as classification training samples.After noise filtering,feature extraction and feature selection process,the support vector machine(SVM)and Linear Discriminant Analysis(LDA)models are used as classifiers for intent recognition,with an accuracy rate of over 97%.The SVM model is selected as the online classifier for less dependence on the number of samples.In the aspect of exoskeleton trajectory planning,analyse the kinematics model of the lower limb exoskeleton and the normal gait trajectory planning was carried out on working space and joint space respectively.The final exoskeleton motion trajectory is planned on the robot joint space.This method can avoid the singular solution of the inverse solution of kinematics model,ensure the continuity of joint speed and acceleration and can be modified presest trajectory quickly and easily by setting a small number of via points.Finally,the online experiment establishes a control mode which combines intention recognition and finite state machine based on the scene,reducing the number of arm movement classification and increasing the recognition rate to over 98%,realizing the stable and autonomous operation of the exoskeleton robot and improving the safety and user experience of human-computer interaction.