基于CNN模型的sEMG信号手势动作识别算法  

Gesture Recognition Algorithm of sEMG Signal Based on CNN Model

作　　者：王逸桐 许成哲[1] Wang Yi-tong;Xu Cheng-zhe(College of Engineering,Yanbian University,Jilin Yanji 133002)

机构地区：[1]延边大学工学院,吉林延吉133002

出　　处：《电子质量》2021年第1期6-9,共4页Electronics Quality

摘　　要：表面肌电信号是利用表面电极然后记录的肌肉运动产生的生物电信号,表面肌电信号可以用来反映神经、肌肉的状态及功能。科学技术日益发展的今天,越来越多的研究人员针对基于sEMG的手部动作的问题进行探讨。截至目前,在体能训练、身体康复训练、医学临床及运动训练等领域取得巨大突破。随着深度学习模型的日益成熟,各种模型对sEMG信号手势动作的准确率有明显提升。该文提出了基于CNN网络的sEMG信号手势动作识别算法,首先是预处理过程,选用2阶巴特沃斯滤波器对sEMG信号预处理过程,利用标准差来滤除无信号段,之后进行数据进行扩充,满足大量实验数据的需求。最后介绍了基于CNN网络的sEMG信号手势动作识别过程,说明了CNN网络的结构以及参数设置。结果显示在Ninapro数据集上的最高准确率为77.33%,该文采用的算法在识别效果上具有良好的效果。Surface electromyography(sEMG)is a bioelectrical signal generated by muscle movement recorded by surface electrodes.SEMG signals can be used to reflect the state and function of nerves and muscles.Nowadays,with the development of science and technology,more and more researchers are discussing the hand movement based on sEMG.So far,great breakthroughs have been made in the fields of physical training,physical rehabilitation training,medical clinic and sports training.With the development of deep learning model,the accuracy of sEMG signal gesture has been improved significantly.This paper presents a gesture recognition algorithm of sEMG signal based on CNN network.Firstly,it introduces the preprocessing process of sEMG signal by using the second-order Butterworth band stop filter,and then the data is expanded.After that,the process of sEMG signal gesture recognition based on CNN network is introduced,and the structure and parameter setting of CNN network are explained.The results show that the highest accuracy rate is 77.33%on ninapro dataset.The algorithm used in this paper has good recognition effect.

关 键 词：CNN模型 深度学习 手势动作 SEMG信号 Ninapro 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术] TP183[自动化与计算机技术—计算机科学与技术]