Research On Continuous Estimation Of Kinematics And Force Based On EMG From Human Upper Limb

Surface electromyography(sEMG)signals are physiological signals coming from human body.These signals can be utilized to continuously estimate human's kinematics intention.Recently,there still exists errors between the measured result and result from the sEMG-based human's continuous kinematics estimation.One of the reasons may be because that during complex tasks,the change of task-related variables,such as moving speed,the posture of limbs and external load,may leads to the change of pattern of sEMG signals.To figure out the influence from different task-related variables will be of help to optimize the experiment protocols and improve the estimation performance.Meanwhile,the sEMG signals are non-stationary,which may contribute to the fluctuation of estimation results.Using better feature extraction process may suppress such phenomenon.Moreover,because of the complexity of activities of daily living(ADL),to estimate only human's kinematics intention is insufficient for compliance control.It is also necessary to estimate the interactive force between human body and the environment.Thus,this thesis purposes some researches to these problems,including:(1)Figure out how much some task-related variables influenced the performance of sEMG-based continuous kinematics estimation.In this research,the flexion of elbow joint is taken as example to analyze the influence caused by the change of moving speed,external load and the posture of upper limb.Analysis results show that the moving speed have significant influence on the estimation performance,while the other two task-related variables have little effect.This conclusion provides suggestions for the design of experiment protocols.(2)Adapt Bayes filter to the preprocessing process of sEMG signals and analyze its performance on continuous kinematics estimation of human movement.The result of experiments show that this method can effectively reduce the fluctuation of the features of sEMG signals.Meanwhile,this method can also improve the stability of estimation results without introducing extra errors.(3)Design and build sEMG-based kinematics and force continuous estimation method.The method is evaluated by estimating drilling-like motion with sEMG signals from upper limb.In the experiment,sEMG preprocessing process based on Bayes filter is used to extract features of sEMG.Non-negative matrix factorization(NMF)is applied to decompose the coupling relationship among the sEMG features from multiple channels.To performance continuous estimation,neural network model is applied to learn the relationship between the decomposed sEMG features and the target values.The experiment results and the analysis illustrate that with the sEMG signals of relative muscles,it is possible and effective to synchronically continuous estimate the human intention of joint position and interactive force with objects.