Research On Human Motion Estimation Based On Block Structure Model

The surface EMG signal of human body contains rich information related to human motion.It has natural advantages to use it as an information interaction medium to build an exoskeleton human-computer interaction system.Identifying human motion intention through EMG signal is a key link to achieve human-computer natural interaction control.However,due to the individual differences of EMG signals and their vulnerability to non ideal conditions such as muscle fatigue in practical applications,it is a challenging task to accurately and robustly identify human motion intentions based on EMG signals.In this dissertation,the continuous motion intention estimation of human body based on EMG signals is studied in depth.The main contents include the continuous estimation of human motion intention,the online supervision and evaluation algorithm of muscle fatigue,and the exoskeleton interactive control system based on fatigue supervision.The details are as follows:First of all,combined with the analysis of muscle activation dynamics,muscle contraction dynamics,musculoskeletal geometric model and joint motion equation,the lower limb neuromuscular skeletal model simplified to obtain joint motion from myoelectric signals was deduced,and the block Hammerstein model was used to fit it.The model was similar to the neuromuscular skeletal model in mathematical expression,and the model structure could also correspond to the neuromuscular activation and muscle contraction of the neuromuscular skeletal model,It is one of the possible structures of biological system identification,and the feasibility and accuracy of the model fitting expression of neuromuscular bone model are verified through human joint motion angle prediction experiment,which lays the model foundation of the full text.Then,aiming at the non ideal condition of muscle fatigue,we designed an experimental paradigm of muscle fatigue trend,established Hammerstein motion model based on EMG signal and joint motion angle,identified and tracked the changes of the model online,explored the rules between the changes of the model and the muscle state,summarized the quantitative design of muscle fatigue evaluation algorithm,and built a feature augmentation matrix by extracting the characteristics of the changes of model parameters,Experimental machine learning hierarchical clustering is used to cluster the degree of fatigue,realizing the method of real-time monitoring and evaluating the degree of muscle fatigue under the dynamic motion state.Through multi angle experiments,it is verified that this algorithm can grade the degree of fatigue on the premise of ensuring the accurate trend,which has important guiding significance for the exoskeleton control system.Finally,an exoskeleton adaptive human-computer interaction control method is proposed based on the online evaluation algorithm of muscle state.On the basis of no additional sensors,the evaluation algorithm is used as the perception and decision-making link of the control system to monitor the muscle state in real time and guide the decision-making of the execution and control link.The damping coefficient of the impedance control model is updated online to achieve adaptive variable damping control,Then the motor carries out mechanical drive and assistance according to the control strategy,and the exoskeleton feeds back the motion information to the perception and decision-making links to form a closed loop.The experimental results show that the human-computer interaction method can adaptively change the assist torque on the premise of ensuring the given reference track,and ensure the safety and efficiency in the use process.The work of this dissertation expands the research content of human-computer interaction technology based on EMG signal,deeply explores how to improve the accuracy and robustness of EMG system,and establishes theoretical and experimental research on the natural compliance control method of exoskeleton robot,which has broad application prospects.