表面肌电与三轴信息融合的运动判断实验  

Motion Judgment Experiment by Using Surface Electromyography and Three-axis Information Fusion

作　　者：喻剑[1,2] 李至霖 庞鹏瞩 李洁 YU Jian;LI Zhilin;PANG Pengzhu;LI Jie(School of Electronics and Information Engineering,Tongji University,Shanghai 201804,China;National Demonstration Center for Experimental Computer and Information Technology Education,Tongji University,Shanghai 201804,China;Shanghai Yangzhi Rehabilitation Hospital(Shanghai Sunshine Rehabilitation Center),Tongji University,Shanghai 201804,China)

机构地区：[1]同济大学电子与信息工程学院,上海201804 [2]同济大学计算机与信息技术国家级实验教学示范中心,上海201804 [3]同济大学上海市养志康复医院(上海市阳光康复中心),上海201804

出　　处：《实验室研究与探索》2024年第3期23-27,共5页Research and Exploration In Laboratory

基　　金：中国残联课题残疾人辅助器具专项(2023CDPFAT-12);上海申康医院发展中心医企融合创新协同专项(SHDC2023CR001)。

摘　　要：为了提高基于表面肌电与三轴加速度信号的运动识别准确率,提出了一套多源信息融合处理的实验流程与方法。该方法利用5层离散小波变换对表面肌电信号进行分解,充分提取不同运动产生的肌电信号中各频域的特征信息;再将分解后的表面肌电信号与三轴加速度信号通过滑动窗口的方法进行特征融合,构造融合肌电与空间运动特征的特征图;最后用融合特征图对深度学习模型进行训练,并结合自动状态机进行最终运动状态的识别。实验结果表明,多源信息融合处理方法可以提高运动识别的准确性,总体识别精度分别达到了95.4%和89.2%。该方法在实时性与准确性上均有良好表现。In order to improve the accuracy of motion recognition based on surface electromyography and three-axis acceleration signals,a set of experimental procedures and methods for multi-source information fusion processing are proposed.Firstly,this experimental method uses five-layer discrete wavelet transform to decompose the surface electromyographic signal and fully extract the characteristic information of each frequency domain in the electromyographic signal generated by different movements.Secondly,the decomposed surface electromyographic signal and the three-axis acceleration signal are combined by sliding window method and construct a feature map that fuses electromyographic and spatial motion features.Finally,the fused features map are used to train the deep learning model,and the trained model is combined with an automatic state machine to identify the final motion state.Experimental results show that the multi-source information fusion processing method can improve the accuracy of motion recognition,with the overall recognition accuracy reaching 95.4%and 89.2%respectively.It has good performance in real-time and accuracy.

关 键 词：多源信息融合 表面肌电信号 运动识别 时频分析 深度学习 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]