The lower Limb exoskeleton robot has a very broad application prospect in assisting,helping the elderly and helping the disabled.At present,the exoskeleton robot has developed more mature in structural creation.However,it is still lacking in man-machine coordination and human body following.Patients can't achieve active and arbitrary control when wearing lower limb exoskeleton for rehabilitation training.The surface Electromyography(sEMG)signal can reflect the degree of contraction of the human muscle about 100 ms in advance.Therefore,the sEMG can be used to predict the human motion intention,and the signal can be used as a control signal source for the exoskeleton robot.In this paper,the real-time gait switching and active rehabilitation training control methods of exoskeleton robots based on sEMG signals are studied for patients with lower limb function impairment during and after rehabilitation.Firstly,the sEMG signal preprocessing method of power-frequency notch + bandpass filter and the feature extraction method of root-mean-square + sliding window + smooth processing were proposed.The extracted results were compared with the joint Angle signals collected synchronously to identify the muscles most related to joint movement.BP neural network algorithm was used to establish a joint angle prediction model based on sEMG signals,and BP neural network transfer function,learning rate and other related parameters were determined.sEMG signals and angle signals of knee and hip joints were collected when knee and hip joints were in motion.The extracted sEMG signals and angle signals were input into BP neural network for training,and a joint angle prediction model based on sEMG signals was obtained.According to the joint prediction angle,the lower extremity exoskeleton robot crosses the obstacle gait switching logic and the exoskeleton active triggering and variable step control strategy.are established.A real-time communication interface between the upper computer,the exoskeleton robot,the myoelectric acquisition device,and the angle acquisition device was established.The muscle sEMG signals and joint angle signals were collected and the joint angle prediction model was trained.Subjects wore exoskeleton robots forexperimental verification of gait switching and active control theory.Experiments show that the control method has better real-time performance and accuracy.The theory and practice show that the application of sEMG signal in the control of exoskeleton robot can facilitate the active rehabilitation training of patients and greatly improve the intelligence level and human-machine synergy performance of the exoskeleton robot. |