Motion Perception And Biodynamic Modeling Based On The Fusion Of Multi-source Information Of Lower Limb

The lower limb exoskeleton robot is an intelligent assistant device which can be worn by people.The research in this field demands a combination of Bionics Technology,Biomedical Technology,Artificial Intelligence Technology,Automatic Control Technology,Data Fusion Technology and many other technologies to develop the intelligent system.It can not only carry the loads instead of people,enhance soldiers' fighting capability,but also can be used to treat patients wiith rehabilitation training according to the planned gait in advance.Therefore,the lower limb exoskeleton robot has a promising prospect of application in rehabilitation,military,disaster assistance field and so on.It has become one of the most popular research fields in recent years.The research on the control information of the lower limb exoskeleton robot is a major branch in this field.Whether it can provide effective and intelligent control information will directly influence the comfort of wearing and the performance of the lower limb exoskeleton robot.The surface EMG(sEMG)signal is the direct response of the brain's dominant consciousness and provides more flexible and intelligent control effect as the control signal of the lower limb exoskeleton robot.This thesis carries out the research on how to acquire the useful motion information based on sEMG fusing with inertial information and plantar pressure of lower limbs,which includes the improvement on the classification and recognition of the human lower limb motion,the prediction about the joint angle based on the sEMG and the biodynamic modeling of the lower limbs.The premise of the lower limb exoskeleton robot functioning well is effectively obtaining the motion patterns and states of the wears.In this dissertation,the data of the sEMG,the inertial information and the plantar pressure of lower limbs is processed by the method of data fusion.According to the periodic feature of gait,the gait phases are obtained.The proposed method of feature extraction and classification based on the gait phases is better than the one based on the gait periods and is verified by the experiments.Besides the recognition of the motion patterns of the human lower limbs,the prediction about the joint angle based on the sEMG and the biodynamic modeling of the lower limbs will further improve the comfort of wearing and the tracing accurateness of the lower limb exoskeleton robot.The data of the sEMG of the major muscles in the human lower limbs and the plantar pressure which are trained by the generalized regression neural network(GRNN)algorithm is used to predict the knee joint's angles.In order to improve the prediction accuracy,the Fruit flies optimization method is employed to optimize the structure parameters of the GRNN algorithm.The lower limb biodynamic modeling is an effective method to accurately evaluate the human body's muscles and the strength of joints and is crucial to the clinical rehabilitation,exoskeleton auxiliary control and other fields.Based on the physiological analysis of lower limb muscles and the introduction of the related knowledge of the Hill muscle model,the OpenSim simulation software is applied for the lower limb biodynamic modeling and the analysis and validation of the joint and muscle forces.