K-means聚类神经网络分类器在睡眠脑电分期中的应用研究

Applied Research of K-means Clustering Neural Network Classifier in the Sleep EEG Staging

利用近似熵的方法对睡眠EEG信号进行分期,但睡眠Ⅲ期和Ⅳ期近似熵值非常接近,靠近似熵值无法区分,对分期结果中的Ⅲ期和Ⅳ期EEG信号进行AR(自回归)建模,作为该段EEG信号的特征属性,利用K-means聚类的神经网络分类器对睡眠Ⅲ期和Ⅳ期进行分期,达到了很好的分期效果.

Because the sleep stage III and IV approximate entropy （ApEn） values are very close, using the approximate entropy method for sleep staging EEG signal, they are not distinguished. AR model of staging results in stage III and IV EEG signal is establisthed as the EEG signal characteristics,and K-means clustering neural network classifier is used for the sleep phase III and IV stage, so a good stage effect is achieved.