| The research and application of muscle force in biomedical engineering is very extensive.The test of muscle force can be applied to many fields,such as biological kinematics,rehabilitation training,clinical decision-making,sports,bionic joint control,involving all levels from scientific research to life,and has great practical value.At this stage,the main method for assessing the health of human muscles is still to observe the degree of joint activity and muscle status by doctors and make subjective judgments based on experience.The accuracy of this diagnosis method is general,and it is subject to subjective influence.The rehabilitation plan formulated is not effective.Therefore,this paper proposes a neural network prediction model of upper limb muscle force based on EMG signal,and applies this prediction method to the quantitative evaluation of the main muscle health of the upper limb.The main research contents are as follows:Firstly,this paper designs the overall scheme of a human upper limb muscle strength tester based on sEMG.By studying and analyzing the characteristics and influencing factors of sEMG signals,a collection strategy for sEMG signals is developed.The target muscles,signal collection equipment,and feature extraction methods for the study are selected,and the overall technical route of the plan is developed.Through the study of various muscle force prediction methods,the selection of key technologies for muscle force prediction is completed,and the BP neural network is ultimately selected to construct a muscle strength prediction model.Secondly,the experimental data collection platform is built to simulate the arm-wrestling scene,and the relevant experimental data collection and the construction of experimental data set are completed.The electromyographic data and six-dimensional force data of the subjects under the specified movements are collected.After a series of preprocessing operations,the correlation between the time-domain characteristics of electromyographic signals and muscle force is analyzed,and appropriate eigenvalues are selected as the training set for the establishment of the prediction model.Then,the classification model of motion intention and muscle force prediction model based on BP neural network are proposed.The preprocessed EMG signal data is used as the input,and the size and direction of the muscle force during the simulation of wrist breaking are used as the output.The validity of the muscle force prediction model is analyzed from the perspective of prediction error using cross validation method.Finally,based on the muscle force prediction model,a muscle force prediction system is designed.By collecting the EMG signal of the subject under the specified action,the size and direction of the predicted muscle force,as well as the activation and contribution of each muscle,can be displayed intuitively through the interface to achieve an objective and effective evaluation of the subject’s muscle health. |
