| Lower limb rehabilitation training technology is an important issue that has been discussed in the domestic academic and medical circles,and it is also a hot topic at home and abroad.The rehabilitation robot is a robotic system for rehabilitation training of the human body(especially individual patients),which can assist or even replace the doctor for physical rehabilitation training.In this paper,based on the needs of clinical rehabilitation applications,the developed lower limb rehabilitation exoskeleton robot and ergolab surface electromyography acquisition system are used as experimental platform,the external skeletal robot lower limb rehabilitation system design,and the surface electromyography signal(sEMG)feature extraction action classification of lower limb movement The research on the continuous motion estimation of the lower limbs and the evaluation system of the lower limbs was carried out.In this paper,an exoskeleton lower limb rehabilitation training robot is designed.In order to predict the human body motion intention by using sEMG,this paper starts from the characteristics of the myoelectric signal of unilateral lower limb motor function in patients with hemiplegia,and uses the sEMG to perform lower limb movement identification to understand the patient.The intention of the movement,so as to plan the training of the predetermined trajectory of the affected limb through the rehabilitation robot.In order to realize the rehabilitation training based on the willingness of the human body,using the extracted sEMG features for lower limb motion recognition,designing a high-performance pattern classifier is the core problem.In this paper,the classification test experiment of a single BP neural network under different sample sets is used to judge the error rate and the number of network training iterations.It is proved that the proposed segmented AR modeling-PCA feature extraction algorithm is better than the traditional autoregressive model.The method is more advantageous.In addition,the disadvantage of the single BP neural network classifier’s recognition rate is unstable.The Adaboost integrated classification idea is introduced into the sEMG’s action pattern recognition.A BP Adaboost integrated classification algorithm that can measure the output information adaptive weighting is proposed,which can improve the surface muscle.The accuracy of the electrical signal on the intent of the lower limbs of the human body.In the estimation of continuous motion of lower extremities,this paper.proposes continuous motion estimation of three joints of lower limbs based on sEMG and LSTM,and improves the real-time application of continuous motion estimation system by principal component analysis(PCA)dimensionality reduction.It proves that LSTM has better accuracy and stability in processing continuous motion estimation of time series EMG signals,and compared with the training results of traditional network-BP and SVM,it obtains higher accuracy of hip,knee and 踝.Estimation of continuous motion of joints.(The root mean square error of hip and knee is 32.99%,17.91%,20.42%compared with SVM network;24.94%,23.13%,and 16.09%compared with BP neural network)In the evaluation of lower limb rehabilitation,in view of the shortcomings of existing clinical evaluation methods,such as limitations of human factors and difficulty in comprehensive quantification,this paper proposes a rehabilitation evaluation system based on comprehensive analysis of myoelectricity and inertial information,and Analytical Hierarchy Process(AHP).Combined with fuzzy comprehensive evaluation(FCE),the lower limb rehabilitation evaluation model based on analytic hierarchy process-fuzzy evaluation(AHPFCE)was constructed,and the lower limb rehabilitation evaluation system based on myoelectric-inertial information was designed and implemented.Objectively and quantitatively evaluate the lower limb motor.function of the patient,and initially verify the feasibility and effectiveness of the lower limb rehabilitation evaluation system. |
PDF Full Text Request