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

• 开发研究与工程应用 • 上一篇    下一篇

基于脑电和肌电多特征的自动睡眠分期方法

吕甜甜1,2,王心醉 2,俞乾 2,于涌 2,蒋蓁 1   

  1. (1.上海大学 机电工程与自动化学院,上海 200072; 2.中国科学院 苏州生物医学工程技术研究所,江苏 苏州 215163)
  • 收稿日期:2016-08-16 出版日期:2017-10-15 发布日期:2017-10-15
  • 作者简介:吕甜甜(1991—),女,硕士研究生,主研方向为生物医学信号处理;王心醉,副研究员、博士;俞乾,助理研究员、硕士;于涌,研究员、博士;蒋蓁,研究员、硕士。
  • 基金资助:

    国家自然科学基金(61433016);苏州市科技计划项目(ZXY201427,ZXY201429)。

Automatic Sleep Staging Method Based on Multi-features of Electroencephalogram and Electromyogram

LÜ Tiantian 1,2,WANG Xinzui 2,YU Qian 2,YU Yong 2,JIANG Zhen 1   

  1. (1.College of Electromechanical Engineering and Automation,Shanghai University,Shanghai 200072,China; 2.Suzhou Institute of Biomedical Engineering and Technology,Chinese Academy of Sciences,Suzhou,Jiangsu 215163,China)
  • Received:2016-08-16 Online:2017-10-15 Published:2017-10-15

摘要:

为实现准确的自动睡眠分期,且满足泛化能力的需求,基于脑电(EEG)和肌电(EMG)多特征,提出一种自动睡眠分期方法。以MIT-BIH多导睡眠数据库中样本的EEG和EMG为分析对象,采用离散小波变换对原始数据进行滤波预处理,提取EEG的α,β,θ,δ节律波和高频成分的能量比,利用样本熵算法提取EEG的非线性特征。将特征参数输入支持向量机分类器中进行样本训练与分类识别。实验结果表明,该方法的分期准确率可以达到92.94%,相比基于EEG的睡眠分期方法平均准确率提高3.96%,交叉验证平均准确率达82.68%,具有较好的泛化能力。

关键词: 睡眠分期, 脑电, 肌电, 离散小波变换, 能量特征, 样本熵, 支持向量机

Abstract:

In order to achieve accurate automatic sleep staging and meet the needs of generalization ability,an automatic sleep staging method based on multi-features of Electroencephalogram(EEG) and Electromyography(EMG) is proposed.EEG and EMG of MIT-BIH polysomnographic database samples are chosen as the analysis object.The Discrete Wavelet Transform(DWT) is used to make filter processing of the raw data.The energy ratio of α,β,θ,δ rhythm waves and the high frequency component from EMG are extracted.The nonlinear characteristics of EEG are also extracted by Sample Entropy(SampEn) algorithm.These feature parameters are inputted Support Vector Machine(SVM) classifier for sample training and classification recognition.Experimental results show that the periodization accuracy of the proposed method reaches 92.94%.The average accuracy raises 3.96% compared with the sleep staging method based on EEG and the average accuracy of cross validation is 82.68%.The proposed method has better generalization ability.

Key words: sleep staging, Electroencephalogram(EEG), Electromyogram(EMG), Discrete Wavelet Transform(DWT), energy feature, sample entropy, Support Vector Machine(SVM)

中图分类号: