基于表面肌电和位姿信息融合的手势动作识别  

Gesture action recognition based on fusion of surface electromyography and pose information

作　　者：杨庆华[1,2] 金圣权 都明宇 王志恒[1,2] YNAG Qinghua;JIN Shengquan;DU Mingyu;WANG Zhiheng(College of Mechanical Engineering,Zhejiang University of Technology,Hangzhou 310023;Key Laboratory of Special Purpose Equipment and Advanced Processing Technology,Ministry of Education and Zhejiang Province,Zhejiang University of Technology,Hangzhou 310023)

机构地区：[1]浙江工业大学机械工程学院,杭州310023 [2]特种装备制造与先进加工技术教育部/浙江省重点实验室,杭州310023

出　　处：《高技术通讯》2023年第12期1295-1302,共8页Chinese High Technology Letters

基　　金：国家重点研发计划(2018YFE0125600);浙江省基础公益研究计划(LGG19E050023)资助项目。

摘　　要：针对仅通过表面肌电信号(sEMG)进行手势识别难以应对复杂手势的问题,提出一种基于表面肌电和位姿信息融合的手势识别方法。通过双阈值方法对信号活动段进行分割,提取表面肌电信号、位姿信号的特征,使用核主成分分析方法(KPCA)对提取特征进行降维融合,使提取特征中的非线性信息得到较好保留,最后通过随机森林(RF)分类器进行分类识别。实验结果显示,该方法对10名受试者的11种不同手势的最佳平均识别率为98.23%,单个动作的识别准确率均在90%以上,验证了提出方法的可靠性。Aiming at the problem of the difficulty of complex gestures recognition just by surface electromyography(sEMG)signals,a gesture recognition method based on the fusion of surface electromyography and pose information is proposed.First,the signal active segments are segmented by the dual-threshold method,and then the features of the surface electromyography signal and the pose signal are extracted.Second,the kernel principal component analysis(KPCA)is used to perform dimensionality reduction and fusion on the extracted features so that the non-linear information in the extracted features is preserved well.Finally,the random forest(RF)classifier is utilized for classification.The experimental results show that the highest average accuracy of the proposed method is 98.23%for 11 different gestures of 10 subjects,and every individual gesture is above 90%,which verifies the reliability of the proposed method.

关 键 词：表面肌电信号(sEMG) 位姿信号 多模融合 核主成分分析方法(KPCA) 随机森林(RF) 

分 类 号：R318[医药卫生—生物医学工程] TN911.7[医药卫生—基础医学]