摘要
目的:因睡眠问题的日益严重,且睡眠分期是睡眠状况分析和睡眠质量评价的前提和基本内容,所以本文的主要目的就是基于脑电信号研究睡眠自动分期的方法。方法:使用小波变换对脑电信号进行预处理,因为脑电信号的非平稳性选择非线性分析方法,提取信号的复杂度和近似熵作为睡眠脑电各个时期的特征值,最后利用支持向量机对睡眠各阶段进行分期决策。结果:睡眠各期脑电的近似熵值和复杂度值随着睡眠状态的变化而不同,睡眠各期可以根据特征值的不同而得到有效区分,通过对1458个脑电信号样本进行自动分期,得到平均准确率为85.67%。结论:小波变换可以很好的对脑电信号的消噪处理,而脑电信号的复杂度值和近似熵值作为特征值,可以作为睡眠分期的有效分类依据。
Objective： Due to sleep problems are growing, and classification for sleep stage is premise and basis of sleep＇s analysis and evaluation. So this paper aims at the auto classification for sleep stage based on electroencephalogram （EEG）. Methods： The original signals were preprocessed by the means of wavelet transform. We chose nonlinear analytical methods because of the nonstationarity of EEG signal. The complexity and approximate entropy were extracted from the denoised sleep data. Finally support vector machine （SVM） was adopted for the auto classification for sleep stage. Results： The value of complexity and approximate entropy were different with the change of sleep state. Sleep stages were effectively distinguished through the test of 1698 samples, and the average accuracy of auto classification was 87.34%. Conclusion： The results show that wavelet transform can play well in the denoising processing for EEG signal. The sleep stages can be effectively determined with the complexity and approximate entropy as feature value.
出处
《软件》
2013年第2期97-100,共4页
Software
关键词
睡眠自动分期
EEG
小波变换
复杂度
近似熵
支持向量机
sleep stage classification
EEG
wavelet transform
complexity
approximate entropy
support vector machine
