基于多普勒雷达的非接触式睡眠监测关键技术研究

发布时间：2017-12-27 23:08

本文关键词：基于多普勒雷达的非接触式睡眠监测关键技术研究　出处：《南京理工大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着现代人生活节奏的加快,越来越多的人们被睡眠障碍及相关疾患所困扰。因此早期进行睡眠监测,观察睡眠分期,研究睡眠规律对睡眠相关疾病的诊断与治疗意义重大。传统睡眠分期方法主要依靠脑电、眼电、肌电和心电进行睡眠分期,该方法存在操作复杂、电极繁多、监测价格昂贵和测试者有心里压力等问题,因此研究非接触式睡眠分期方法具有重要意义。基于多普勒雷达的睡眠监测具有非接触、操作简单、能够长期监测等优点,有潜力应用于家庭日常睡眠监测。本文针对基于多普勒雷达的非接触式睡眠监测展开了相关研究,其主要工作如下:1、介绍了脑电信号的形成、特点及其特征波段与基于脑电信号睡眠分期标准与分期方法。2、研究了多普勒雷达基本原理与非接触睡眠监测系统,根据生理信号特点设计滤波器,从雷达回波中提取呼吸和心跳信号,根据静息状态下的雷达回波特点从中提取测试者睡眠时的体动信号。通过波谷检测法提取呼吸和心跳次数,模拟睡眠状态,验证了非接触式睡眠监测系统在长时间监测时系统稳定并且所测呼吸、心跳和体动信号准确。3、提出基于决策树睡眠分期算法,从呼吸和体动信号中提取不同睡眠阶段的6个特征参数,通过特征参数与不同睡眠阶段的对应关系,建立睡眠分期决策树,选取特征参数取值范围,完成睡眠分期。实测数据实验表明所提算法可达到59.46%的平均分期精度。4、提出基于支持向量机睡眠分期算法,从呼吸、心跳和体动信号中提取不同睡眠阶段的11个特征参数,将输入信号分为有/无心跳信号特征参数,优化参数选择,构建支持向量机模型,完成睡眠分期。实测数据实验表明,支持向量机睡眠分期算法可达72.84%的平均分期精度,验证了所提算法的有效性。
[Abstract]:With the rapid pace of modern people's life, more and more people are troubled by sleep disorders and related diseases. Therefore, early sleep monitoring, observation of sleep stages, and the study of sleep patterns are of great significance for the diagnosis and treatment of sleep related diseases. The traditional sleep staging method mainly depends on EEG, EOG, EMG and ECG of sleep stages, the method has complicated operation, electrode variety, and test monitoring expensive psychological pressure and other issues, so it is important to research the non-contact sleep staging method. The sleep monitoring based on Doppler radar has the advantages of non-contact, simple operation and long-term monitoring, and has the potential to be applied to daily sleep monitoring in the family. In this paper, we carried out related research on non-contact sleep monitoring based on Doppler radar. Its main works are as follows: 1, we introduced the formation, characteristics and characteristics of EEG, and the staging and staging methods based on EEG sleep staging. 2, we studied the basic principle of Doppler radar and the non-contact sleep monitoring system. According to the characteristics of physiological signals, we designed filters, extracted respiratory and heartbeat signals from radar echoes, and extracted the body moving signals of the tester during sleep based on the characteristics of radar echo in resting state. Through the trough detection method, the number of breaths and heartbeats was extracted, and the sleep state was simulated. It verified the stability of the non-contact sleep monitoring system during long time monitoring and the accuracy of the measured respiration, heartbeat and body movement signals. 3, the proposed decision tree algorithm based on sleep staging, extracted from the 6 characteristic parameters of different stages of sleep breathing and body movement signals, corresponding to the relationship between the characteristic parameters and the different stages of sleep, a sleep staging decision tree, select the feature parameters of complete sleep period. The experimental data show that the proposed algorithm can reach 59.46% of the average staging accuracy. 4, the support vector machine algorithm based on sleep staging, extracted from the 11 characteristic parameters of different stages of sleep signals breathing, heartbeat and body movement, the input signal is divided into / out of the heartbeat signal characteristic parameters, optimization parameter selection, construction of support vector machine model, complete sleep staging. The experimental data show that the support vector machine sleep stage algorithm can reach 72.84% of the average staging accuracy, which verifies the effectiveness of the proposed algorithm.
【学位授予单位】：南京理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R740;TN911.7

【相似文献】