摘要
针对现存基于肌电信号的动作模式识别方法数据量不足、特征融合冗余、分类器识别精度低、泛化能力差且动作类别少等问题,以下肢动作为研究对象,采集爬坡、平地行走、上楼以及下楼4种动作类别的表面肌电信号(surface electromyography,sEMG),提出一种基于特征相关性和任务贡献度的特征筛选方法,最终实现了多特征融合的下肢动作模式识别。该方法在提高下肢动作模式识别的效率与精度方面显著优于传统的单特征和原始信号识别方法,可为特征筛选、多特征融合动作模式识别研究提供参考。
Existing action pattern recognition methods based on electromyography(EMG)signals have challenges such as insufficient data volume,redundant feature fusion,low classifier recognition accuracy,poor generalization ability,and a limited number of recognized action categories.This study focuses on lower limb movement,collecting surface electromyography(sEMG)signals for four movement categories:walking uphill,walking horizontally,ascending stairs,and descending stairs.This method adopts a feature selection method based on feature correlation and task contribution and finally achieves multi-feature fusion for lower limb action pattern recognition.This method is significantly better than traditional single feature and original signal recognition methods.It provides valuable insights into the study of feature selection and multi-feature fusion in action pattern recognition.
作者
黄重宇
HUANG Chongyu(School of Electromechanical and Vehicle Engineering,Chongqing Jiaotong University,Chongqing 400074,China)
出处
《技术与市场》
2024年第3期48-52,共5页
Technology and Market
关键词
表面肌电信号
模式识别
特征筛选
特征融合
surface electromyography
pattern recognition
feature screening
feature fusion
