基于心电与呼吸信号的睡眠分期算法研究  

The study of sleep stage classification based on ECG and respiratory signal

作　　者：黄文汉[1] 张伟[1] 胡立刚[1] 周陈旺 向丹阳 

机构地区：[1]同济大学电子与信息工程学院,上海201804

出　　处：《智能计算机与应用》2018年第1期49-54,共6页Intelligent Computer and Applications

摘　　要：睡眠是人类的一项重要生理活动,其质量与人体健康状态密切相关,对睡眠情况进行分析有助于许多疾病的预防和监测。传统的睡眠分期黄金标准是多导睡眠图,并包含脑电、眼电、肌电、呼吸、血氧、运动等多种信号,操作复杂且对测试者睡眠有影响。因此,基于可穿戴设备和有限类型数据进行睡眠分即已成为当前研究热点。本文仅采用了心电和呼吸信号进行特征提取,使用前向序列选择方法进行特征选择,分别采用支持向量机、随机森林和AdaBoost等方法进行分类,在睡眠呼吸障碍患者数据库中进行WAKE-REM-NREM分期上获得最高71.9%的准确率(Kappa=0.36)。实验表明心电与呼吸信号可在一定程度上代替多导睡眠仪应用于睡眠呼吸障碍患者的分析,有助于睡眠呼吸类疾病的诊断和评价,为相关设备的便携化提供了算法基础。Sleep is an important physiological activity of human body. Its quality is closely related to human health status.Analyzing sleep conditions can help to find out prevention and monitoring of many diseases. The traditional gold standard of sleepstage classification is polysomnography, which contains various signals such as EEG, EOG, EMG, respiration signal, bloodoxygen, body movement, etc. But its disadvantages are complicated to operate and having an impact on the testers＇ sleep. Therefore,sleep stage classification based on wearable devices and limited types of data has become the current research hotspot. This paperadopts ECG and respiratory signal to extract features, uses forward sequence selection method to select features. The classifier adoptsSVM, Random Forest and AdaBoost respectively. The accuracy of classification of WAKE-REM-NREM is up to 71.9% （Kappa =0.36） in sleep apnea database. Experiments have shown that ECG and respiratory signals can be used to analyze sleep condition ofsleep apnea patients instead of polysomnography, which contribute to the diagnosis and evaluation of sleep-breathing diseases, andprovide an algorithm basis for the portable devices.

关 键 词：睡眠分期 心电信号 呼吸信号 特征提取 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]