基于黎曼普鲁克的手部离散动作识别方法  

Grasp motion pattern recognition on Riemann Procrustes analysis

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

机构地区：[1]浙江工业大学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 [2]浙江工业大学机械工程学院,杭州310023

出　　处：《高技术通讯》2024年第8期854-863,共10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：肌电信号能反映人体的运动意图,是外骨骼和假肢控制的主要信号之一。但受试者间的差异,增加了基于表面肌电信号(sEMG)的手部离散动作识别使用成本。针对这一情况,本文从域适应的角度出发,提出一种基于小型调整集的迁移学习建模方法。该方法利用黎曼普鲁克分析(RPA)提取黎曼特征与传统时域特征作为支持向量机(SVM)的输入特征,并通过实验验证了其识别精度。在10名受试者身上进行了实验,在黎曼特征下黎曼普鲁克分析相比于不进行迁移学习的动作识别方法提高了5%~7%的准确率。在特征空间分布上,黎曼普鲁克分析后的黎曼特征的重合度更高。结果表明,该方法在基于肌电信号的手部离散动作识别上有明显优势。Electromyography(EMG)can reflect the movement intention of the human and it is one of the main signals for exoskeleton and prosthetic control.However,inter-subject variability increases the cost of using surface electromyography(sEMG)-based discrete hand motion recognition.In response to this situation,a transfer learning method is proposed based on small adjustment sets from the perspective of domain adaptation.This method utilizes Riemann Procrustes analysis(RPA)to extract Riemannian features and traditional time-domain features as the input features of support vector machine(SVM),and its recognition accuracy is verified by experiments.Experiments are carried out on ten subjects,and the Riemann-Plucker analysis under the Riemann feature increases the accuracy by 5%to 7%,compared with the action recognition method without transfer learning.In terms of feature space distribution,the overlap of Riemannian features after Riemann-Plucker analysis is higher.The results show that this method has obvious advantages in recognition of discrete hand movements based on EMG signals.

关 键 词：表面肌电信号(sEMG) 黎曼普鲁克分析(RPA) 手势识别 支持向量机(SVM) 迁移学习 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TP18[自动化与计算机技术—计算机科学与技术]