改进PSO优化LSTM的表面肌肉电信号手势识别方法  

Research on rehabilitation robot system based on sEMG gesture recognition

作　　者：戴坚怿 路光达[1,2] 秦转萍[1,2] 刘欣霖 郭庭航 殷芊涵[1,2] DAI Jianyi;LU Guangda;QIN Zhuanping;LIU Xinlin;GUO Tinghang;YIN Qianhan(School of Automation and Electrical Engineering,Tianjin University of Technology and Education,Tianjin 300222,China;Tianjin Key Laboratory of Information Sensing and Intelligent Control,Tianjin University of Technology and Education,Tianjin 300222,China)

机构地区：[1]天津职业技术师范大学自动化与电气工程学院,天津300222 [2]天津职业技术师范大学天津市信息传感与智能控制重点实验室,天津300222

出　　处：《天津职业技术师范大学学报》2024年第4期8-12,共5页Journal of Tianjin University of Technology and Education

基　　金：天津市教委科研计划项目(2022ZD035)。

摘　　要：表面肌肉电信号能快速准确地反映人体的运动意图,通过表面肌肉电信号进行手势识别在康复机器人人机交互领域有广泛研究。针对目前使用长短期记忆(LSTM)神经网络的手势识别算法分类准确率低,粒子群优化(PSO)算法存在全局搜索能力不足、易陷入局部最优的问题,研究了基于改进粒子群优化长短期记忆(IPSO-LSTM)神经网络的表面肌肉电信号手势识别方法。使用主成分分析(PCA)方法对表面肌肉电信号特征进行选择,选择贡献率高的特征组合作为特征空间,并作为IPSO-LSTM模型的输入;采用线性递减惯性权重和非对称学习因子对PSO进行改进,增加PSO算法对全局的搜索能力。将IPSO-LSTM模型与粒子群优化长短期记忆(PSO-LSTM)神经网络和长短期记忆神经网络进行识别准确率的实验对比。实验结果表明:IPSO-LSTM模型的平均识别准确率为96.9%,比PSO-LSTM和LSTM的识别准确率分别提高了7.0%和2.9%,在表面肌电(sEMG)信号手势识别分类中采用IPSOLSTM模型具有优势。Surface electromyogram(sEMG) signals are instrumental in the rapid and precise detection of human motion in tents,making them a focal point in the research of human-robot interaction for rehabilitation robots.To address the suboptimal classification performance of long short-term memory(LSTM) neural networks in gesture recognition tasks,as well as the inherent limitations of particle swarm optimization(PSO) including a lack of robust global search capabilities and a tendency to converge prematurely on local optima,we investigated an sEMG-based gesture recognition method using the Improved Particle Swarm Optimization-Long Short-Term Memory(IPSO-LSTM) neural network.Principal component analysis(PCA) was employed to select the sEMG signal features,and feature combinations with high contribution rates were selected as the feature space and input into the IPSO-LSTM model.The PSO was improved by using linearly decreasing inertia weights and asymmetric learning factors to enhance the global search capability of the PSO algorithm.The IPSO-LSTM algorithm was compared with Particle Swarm Optimization-Long Short-Term Memory(PSO-LSTM) and LSTM.Experimental results show that the average recognition accuracy of IPSO-LSTM model is 96.9%,which is 7.0% and 2.9% higher than that of PSO-LSTM and LSTM algorithms,respectively.The IPSO-LSTM model demonstrates advantages in gesture recognition and classification using sEMG signals.

关 键 词：手势识别 表面肌电(sEMG)信号 粒子群优化算法(PSO) 主成分分析(PCA) 

分 类 号：R31[医药卫生—基础医学] TN911.7[电子电信—通信与信息系统]