作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程 ›› 2021, Vol. 47 ›› Issue (5): 273-276,284. doi: 10.19678/j.issn.1000-3428.0057923

• 开发研究与工程应用 • 上一篇    下一篇

基于统计聚类方法的儿童下肢肌电信号周期识别

闫成起1, 赵利华2, 陈梦婕2, 周军1   

  1. 1. 上海交通大学 电子信息与电气工程学院 电子系, 上海 200240;
    2. 上海交通大学医学院附属儿童医院, 上海 200062
  • 收稿日期:2020-03-31 修回日期:2020-05-28 发布日期:2020-05-20
  • 作者简介:闫成起(1994-),男,硕士研究生,主研方向为统计机器学习;赵利华、陈梦婕,医师;周军,副教授。
  • 基金资助:
    上海交通大学“科技创新专项资金”(YG2017MS33)。

Period Identification for Electromyography Signals of Children’s Lower Limb Based on Statistical Clustering Method

YAN Chengqi1, ZHAO Lihua2, CHEN Mengjie2, ZHOU Jun1   

  1. 1. Department of Electronics, College of Electronic Information and Electrical Engineering, Shanghai Jiao Tong University, Shanghai 200240, China;
    2. Children's Hospital Affiliated to Medical College of Shanghai Jiao Tong University, Shanghai 200062, China
  • Received:2020-03-31 Revised:2020-05-28 Published:2020-05-20

摘要: 为运用肌电信号分析髋脱位儿童和正常儿童的差异,提出一种基于统计的聚类方法,识别步态中下肢肌电信号的周期起始时刻。使用非参数贝叶斯模型将肌电信号序列聚类为状态序列,并通过k均值聚类算法将该状态序列标记为肌肉活跃和不活跃两种状态,将肌肉活跃状态的起始时刻作为肌电信号周期的起始位置,并且利用窗函数方法提高预测准确性。实验结果表明,该方法对于预测正常儿童周期起始位置的识别误差较小,平均值为2.15%,并且在5%的置信度水平下与SampEN、SNEO和IP等检测算法相比具有较高的预测准确率。

关键词: 肌电信号, 周期识别, 统计聚类方法, 非参数贝叶斯模型, k-means算法, 滑动窗

Abstract: To promote the application of Electromyography(EMG) signals in the analysis of differences between normal children and children with hip dislocation,this paper proposes a method based on statistical clustering for detecting the starting point of the period of EMG signals from lower limb mulscles of walking children.The method employs the nonparametric Bayesian model to cluster EMG signal sequences as pattern sequences,which are subsequently marked with tags of active state and inactive state by using the k-means algorithm.The starting point of the active state of muscle activities is taken as the starting point of a period of EMG signals,and the window function method is used to improve the prediction accuracy.Experimental results show that the average recognition error of this method is as small as 2.15%,and is significantly different from that of the other detection algorithms,including SampEN,SNEO and IP when the confidence level is 5%.

Key words: Electromyography(EMG) signal, periodic identification, statistical clustering method, nonparametric Bayesian model, k-means algorithm, sliding window

中图分类号: