摘要
睡眠是一种重要的生理现象,对睡眠进行合理分期,是研究睡眠质量、诊断睡眠疾病的基础。脑电是睡眠过程中最显著和直观的信号,也是研究睡眠的重要且有力的工具。本研究提取了多种脑电相关特征作为识别睡眠脑电信号的指标,并采用多元逐步回归分析法进行特征筛选,通过线性分类及支持向量机(SVM)算法实现了脑电睡眠分期的自动判别。实际测试结果表明,基于单路脑电的睡眠分期判别方法的平均正确率为78.85%,说明该方法较为准确。
Sleep is an important physiological phenomenon and sleep staging is the basis for evaluation of sleep quality and diagnosis of sleep diseases. As the most significant and intuitive signal, EEG(Electroencephalograph) signal has been widely used in sleep studies. In this paper, a variety of EEG correlation characteristics were extracted to identify sleep EEG signals and feature selection was performed by the multiple stepwise regression method. Then sleep staging was realized by linear classification and support vector machine(SVM) algorithm. According to the actual test results, the average accuracy of the sleep staging method based on single channel EEG signal was 78.85%, which proved the effectiveness of the method.
出处
《中国医疗设备》
2015年第12期34-37,共4页
China Medical Devices
基金
国家自然科学基金(7151101018)
北京市日新人才(015000514115006)
关键词
睡眠监护系统
睡眠分期
单路脑电
多元逐步回归分析
线性分类
支持向量机
sleep staging
single channel electroencephalograph
multiple stepwise regression analysis
linear classification
support vector machine