Research On The Relationship Between SEMG And Typical Leg Movements Based On BP Neural Network

SEMG signal is the bioelectrical signal collected on the skin surface when the human muscle is active.It can reflect the activity state of muscle and nerve,to a certain extent,the movement state of human body.Because of the convenient extraction method,it is widely used in the field of human-computer interaction with rehabilitation medicine and intelligent robot.With the aging of the global society,the proportion of people with mobility difficulties is increasing.In order to assist the daily life of people with mobility difficulties,the lower limb exoskeleton robot and other sports auxiliary equipment are growing,and the use of motion recognition technology makes the sports auxiliary equipment more intelligent.At present,there are some problems in the users'intention recognition,such as inaccuracy,cumbersome wearing and redundant signal acquisition.In view of the above problems,this paper will use neural network to recognize the surface EMG signals of four typical movements of the lower limbs,to achieve a higher recognition level with the minimum number of muscles,to reduce the burden of users and improve the utilization rate.The main work of this paper is as follows:1.Analyze the generation mechanism and characteristics of sEMG signal,select and determine the acquisition sensor and master controller,and build the acquisition system.2.According to the anatomical knowledge,analyze and determine the lower limb muscles related to movement.And complete the acquisition test of surface EMG signals of four typical movements of lower limbs.3.Analyze the original signal,build Butterworth filter for noise,preprocess and analyze the collected signal.4.Analyze the filtered surface EMG signal,analyze and extract the features in time and frequency domain.BP neural network is constructed,and the extracted eigenvalues are used as input to classify the pattern actions.And compared the optimized BP neural network model with the input quantity.It is concluded that the recognition rate is the highest when the three-layer BP neural network with 12 neurons is used for the four types of lower limb movement recognition,and the RMS,median frequency and average power frequency are used as the eigenvector input.The research of this paper has important reference value for the application of sEMG signal and the development of recognition system of lower extremity exoskeleton robot.