基于心电脉搏信号的睡意检测方法研究

发布时间：2018-08-31 20:32

【摘要】：随着现代生活节奏的加快，睡眠不足、睡眠不良，甚至彻夜失眠的情况时有发生。人们在追求高效生活和工作的同时，也受到疲劳和睡意的危害。在高危操作环境下，像驾驶、钢铁领域以及核工业等，如果操作人员在工作时有睡意，那人们的生命财产安全都会受到威胁。而在一些病患的状态监测方面，比如阻塞性睡眠呼吸暂停综合征，实时的睡意监测也是十分必要的。心电和脉搏信号蕴含丰富的人体生理信息，测量也较为简便。研究过程中设计了睡意检测实验并且同步采集了30例健康无睡眠问题的实验对象在清醒和睡意状态下的心电信号、脉搏信号和脑电信号。本文首先对筛选出的26例实验对象的心电和脉搏同步信号进行预处理，再通过特征提取与分析，选择在清醒和睡意状态下有显著性差异的特征用于分类。与清醒状态相比，在睡意状态下心电T波峰值、重搏波高度和脉搏波传输时间均非常显著地降低(p0.005)，脉搏波起点到主波峰点的时间非常显著地增大(p0.005)，心率值和心率变异性中的VLF值显著地降低(p0.05)。使用线性判别分析和支持向量机两种方法对这些特征分别进行单一特征和组合特征的分类。本文除了提取到心电和脉搏信号在时域和频域上的20多种特征，计算了心电信号的小波能谱熵，还对睡眠剥夺的实验对象连续时间段内的心电RR间期最值和脉搏波特征K值进行了分析。研究结果表明：睡意对心电和脉搏信号的一些特征有影响；心电T波峰值、重搏波高度、脉搏波传输时间、脉搏波起点到主波峰点的时间、心率以及心率变异性中的VLF在清醒和睡意两种状态下有显著性差异；特征组合有利于提高分类正确率；心电T波峰值的单一特征及其参与的组合特征的分类正确率高于其他。心电和脉搏特征能够用于区分清醒和睡意状态，而且心电和脉搏信号的测量方便，为睡意检测提供一种新的识别方法。
[Abstract]:As the pace of modern life accelerates, sleep deprivation, poor sleep, and even sleeplessness happen all night. People in the pursuit of efficient life and work, but also by fatigue and sleep harm. In high-risk environments, such as driving, steel and the nuclear industry, people's lives and property are threatened if they are sleepy at work. In some patients, such as obstructive sleep apnea syndrome, real-time sleep monitoring is also necessary. ECG and pulse signals contain abundant physiological information of human body, and the measurement is also relatively simple. In the course of the study, a sleepiness detection experiment was designed and 30 healthy subjects without sleep problems were simultaneously collected ECG signals, pulse signals and EEG signals in awake and sleepy states. In this paper, the electrocardiogram and pulse synchronization signals of 26 selected subjects were preprocessed, and then, by feature extraction and analysis, the significant differences between awake and sleepy state were selected for classification. Compared with awake state, the peak value of ECG T wave in sleeping state, The height of repulse wave and the transmission time of pulse wave decreased significantly (p0.005), the time from the starting point of pulse wave to the peak point of main wave increased significantly (p0.005), and the VLF value of heart rate and heart rate variability decreased significantly (p0.05). Linear discriminant analysis (LDA) and support vector machine (SVM) are used to classify these features into single feature and combined feature respectively. In this paper, in addition to extracting more than 20 characteristics of ECG and pulse signals in time and frequency domain, the wavelet energy spectrum entropy of ECG signal is calculated. The maximum value of ECG RR interval and characteristic K value of pulse wave in the continuous period of sleep deprivation were also analyzed. The results show that sleep affects some characteristics of ECG and pulse signal, the peak value of ECG T wave, the height of repulse wave, the time of pulse wave transmission, the time of starting point of pulse wave to the peak point of main wave, VLF in heart rate and heart rate variability was significantly different between awake and sleepiness states; the combination of features was helpful to improve the classification accuracy; the single feature of ECG T wave peak value and the classification accuracy rate of its involved combination characteristics were higher than those of others. The characteristics of ECG and pulse can be used to distinguish awake and sleepy states, and the measurement of ECG and pulse signals is convenient, which provides a new recognition method for sleepiness detection.
【学位授予单位】：兰州理工大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：R318.0;TN911.7

【参考文献】