Design Of Intelligent Recognition System For Leg Fatigue Based On SEMG

As the heavy-duty manual labor gradually shifts from manpower to machine work,the various functions of the human body are diminishing,and the decline in leg strength is most obvious.The exercise of the leg muscles is particularly important.In exercise,the grasp of changes in the state of the leg muscles helps to improve the scientific nature of exercise,and the state of muscle fatigue is one of the important reference factors for evaluating muscle function.This paper designs and produces a complete intelligent identification system based on the sEMG of multiple parts of the leg to evaluate the fatigue state of the leg muscles.The sEMG of the 6 target muscles in the leg is picked up by the button feedback electrode piece,and then amplified by the hardware conditioning circuit formed by the AD620 and OP07 chips.The STM32F103C8T6 internal ADC based on the Cortex-M3 core converts the analog signal output from the 6-channel conditioning circuit into a digital signal,and forwards the data packet to the AM335 X development board intelligent terminal based on the ARM Cortex-A8 core through the master-slave Bluetooth module HC-05.The intelligent terminal can realize EMG drawing,wirelessly forward data packets to the server,and display the function of returning the fatigue state value.The server loads the recognition model of the training completion,processes the data stream received in real time,and obtains the evaluation result and sends it to the intelligent terminal,promptly reminding the user to relax the leg muscles,thereby ensuring exercise safety.Through the research on the mechanism and characteristics of myoelectric signal generation,the advantages and disadvantages of various analytical methods are demonstrated.An attempt is made to establish a multi-layer bidirectional LSTM network model that can be combined with multi-site sEMG to analyze the fatigue state of the leg.In the model training experiment,the sEMG data set of 10 acquisition objects in the high-intensity running movement was obtained,and the fatigue state label of the data set was determined according to the running distance of 0-500 m,500-800 m,800-1200 m,1200-1500 m.The data set is pre-processed to obtain the fatigue instance set suitable for the model training input.As an input,the leg muscle fatigue state recognition model based on multi-layer bidirectional LSTM is established and trained.The experimental results on the test set show that the model can perform high-precision state recognition on the fatigue instance set,and the accurate recognition rate reaches 77.34%.In the actual running application,the quantitative assessment method is used to test the discriminative accuracy and generalization universal ability of the whole fatigue state intelligent recognition system.The subjects were divided into two groups,A and B.Among them,the A group of subjects participated in the data set acquisition experiment when the training model was involved,and the B group was reversed.The experimental results show that the accuracy rate of group A is 68%,and the accuracy rate of group B is 52.67%.After analysis,it is concluded that the overall recognition system works stably in running,and the recognition ability and generalization ability of leg fatigue state are not good.The discriminating accuracy at the transition section of fatigue state is greatly reduced,which is a reason for the overall recognition accuracy.After the fatigue state is stabilized,the discrimination accuracy is improved.In addition,the system transmission delay also has a certain impact on the overall recognition accuracy.