| Surface electromyography signal(sEMG)is a comprehensive result of the superimposition of action potentials generated by muscle fiber motor units on the skin surface during muscle contraction.It is a weak electrophysiological signal that contains rich information related to limb movement.Because of its easy-to-collect,safe,non-invasive,and natural-rich features,it has been widely researched and applied in the fields of intelligent prosthesis,rehabilitation training,disease diagnosis,and human-computer interaction in recent years.Precisely decoding force using sEMG is not only helpful to further reveal the physiological mechanism of the human motion system,but also is an important step to realize natural human-computer interaction.However,in the current human-computer interaction equipments that use sEMG signal as the source of control information and EMG recognition technology,most of the research results aim at specific movement recognition,ignoring force decode during human-computer interaction process,and resulting in lack of precise control of force pattern and force strength.At the same time,limited by the performance of synchronous acquisition equipment for sEMG and force signal,the randomness,weakness and susceptibility of sEMG signal,and the complex physiological structure of the human body,force decode based on sEMG faces many challenges.For the above problems,the paper uses sEMG signal to perceive different force patterns and magnitudes of human hand,and conducts theoretical and experimental research from signal detection,signal processing,model construction and application development.The main research content and results are summarized as follows:(1)In order to obtain high-quality sEMG signal,extend the type of sEMG signal,improve the robustness of the signal acquisition system and solve the problem of the lack of output of the sEMG envelope signal in the existing sEMG acquisition equipment,the paper uses dry metal electrodes to develop the wearable sEMG sensor with the function of synchronous output of raw sEMG signal and sEMG envelope signal.Not only can tiny muscle contractions be captured,but it also overcomes the problems that one-time electrode stickers can’t be reused,it is inconvenient to wear,and the conductive gel is easy to dry out,which makes the conductivity worse.At the same time,a mechanical signal acquisition system with human grip and pinch force is designed based on the mechanical sensing device,and the force measurement accuracy is ±0.1N.(2)For the problems that the sEMG signal has uncertainty,weakness,and serious noise interference,the paper proposes an adaptive filtering method in the preprocessing of the raw sEMG signal,which overcomes the defects of the traditional digital filtering method that makes the useful information be lost while suppressing the power line interference and the harmonic components generated by the power line interference can’t be suppressed.Due to fewer researches on the preprocessing of the sEMG envelope signal,the paper studies the digital filtering method,the median filtering method,the moving average filtering method,the Savitzky-Golay filtering method and the wavelet transform filtering method,and analyzes the performance of each method by experiments.In order to judge the state of muscle contraction,an activity segment detection method based on short-time Fourier transform and an activity segment detection method based on force signal are proposed,and the performance of each method is analyzed and compared by the experimental results.(3)Linear grip force is the simplest and most direct way of force control.For controlled research on linear grip force is neglected in existing studies.In this paper,sEMG signal is used to implement recognition mechanism research of linear grip force.Taking five kinds of linear grip force as the research object,a classification model of linear grip force is designed based on supervised learning.A method for constructing a force decode model based on geometrical ideas is proposed,which realizes the simulation of different linear grip forces over time by using sEMG signal,and provides a new idea for using sEMG signal to precisely control linear grip force.(4)Current researchers tend to regard the sEMG signal during the process of increasing force and decreasing force as exactly the same,and construct the same sEMG-to-force estimation model.For the problem,the paper analyzes and studies the EMG difference in the process of increasing force and decreasing force.A performance evaluation and screening method for EMG features based on Pearson correlation is proposed.Through the signal equidistant segmentation processing technique,a method for analyzing the difference of EMG based on the statistical T-test is proposed,and the variation rule of EMG difference in the process of increasing force and decreasing force under different muscle contraction levels is obtained according to the experimental result.An estimation model from discrete force to continuous force is constructed,which provides a reference for accurate estimation of force using sEMG signal.(5)The pinch force of the human hand is the effect of the synergy of different fingers,involving complex and delicate finger movement and force control,so the decode of pinch force based on sEMG is still a difficult problem.For the problem,the paper studies six common pinch forces using sEMG.For the problem of the same increase and decrease patterns of feature sequences,a feature sequence similarity evaluation method based on Manhattan distance is proposed,which overcomes the similarity between features.In order to realize the prediction of the pinch force strength and the classification of the pinch force movement at the same time,by using the isometric sample sequence segmentation method,the paper realizes the prediction of the pinch force strength based on the regression model.On above basis,a method of converting the isometric multi-dimensional EMG feature sequence into a normalized grayscale image is proposed,and the classification of different pinch force patterns is realized by using the image recognition method.(6)A multifunctional rehabilitation training system integrating muscle force rehabilitation training,grip force rehabilitation training and pinch force rehabilitation training is developed.Traditional rehabilitation training method is still mainly mechanical and boring exercise,which lacks interesting training process and can’t coordinate the patients’ multiple senses,and it can’t improve the patients’ subjective initiative.For the problem,the paper uses the self-developed signal acquisition hardware system,the proposed signal processing method and specific signal encoding algorithm to encode the sEMG signal,grip force signal and pinch force signal into the control instructions of the visualized object,and turns rehabilitation training into an interesting scene through interesting software design.In the training process,this system can synchronously coordinate the patient’s limbs,eyes,brain and other multiple senses and exercise their synchronous coordination ability and reaction,and maximize the patient’s subjective initiative,and it provides a new form for rehabilitation training. |
PDF Full Text Request