Research On Bioelectric Signal Control Of Upper-Limb Exoskeleton For Movement Intention Recognition

Exoskeleton robot is a research object that is highly interdisciplinary.It has demonstrated the ability to assist people in aerospace,medical,life and industrial applications.In addition to basic functions,the design and control of system need to fully consider some factors such as human safety,natural operation,predictive predictability,system efficacy,operator acceptability and active participation.This is different from the traditional industrial robots.The optimal interaction between people and robot is that the exoskeleton can assist people to perform exercises simply and naturally without any redundant communication with the robot.Motion execution must be identified and activated by appropriate sensors.The interface should be as close to the human nervous system as possible.For the problem of existing exoskeleton,which lack a full understanding of users' motor intentions and limit control performance,the dissertation presented bioelectrical signals to control upper-limb exoskeleton for intentional recognition.The signal was processed by classification or regression to drive the exoskeleton and continuous motion was fully supported.The research focused on the motion identification accuracy of sEMG control and the feasibility of predicting motor intent using EEG Then the theoretical method was verified by exoskeleton design and experiment.The main contents of this dissertation are as following:(1)Biological electric signal control theory for motion intention identification.The control scheme of the biological electrical signal and the pattern of motion intention recognition were studied.Then the advantages and disadvantages of sEMG and EEG were compared in interactive control.And the theory of intention recognition control of the two kinds of signals was researched,which provided the methodological support for the later research.(2)Continuous recognition of wrist motion based on sEMG.Two methods of intention recognition of wrist were studied,including continuous parameter identification and continuous classification recognition.In order to apply the continuous proportional control of sEMG to wrist,the continuous motion of angle parameters was studied in four motion directions.Constant torque and grip force were compared to improve the accuracy of angle recognition.In the continuous classification study,the four categories of wrist movement were identified and the reference results for real-time control were obtained.(3)Effect of interaction factors on sEMG control of elbow joint.First,bioelectrical sensors are more sensitive to the pressure and position of the physical interface than physical sensors.Influence of pressure and position factors on the interface of the elbow joint with the sEMG signal was studied.Second,the effects of internal and external focus and ontology perception(no visual feedback)on the prediction of sEMG were studied by analysis of brain-electromyography coherence and setting of experimental tasks.It is more automatic in motion control and more efficient in motion mode to choose the psychological conditions.(4)Feasibility study on upper limb motion intention recognition based on EEG signal.Depth convolutional neural network algorithm was used to decode EEG signal,and discussed the robustness and accuracy of EEG data prediction for multi-classification of upper limb movements.(5)Application of intention recognition method for upper-limb exoskeleton.Upper limb exoskeleton system was designed,including mechanical design and sEMG signal control design.The scheme of motion intention recognition was proposed for different application fields.And the continuous regression of sEMG signal-angle parameter or the continuous classification of sEMG signal-direction was realized.