基于智能手机的睡眠周期识别技术和应用研究

发布时间：2018-06-23 22:16

本文选题：普适计算 + 智能手机　；参考：《南京大学》2017年硕士论文

【摘要】：睡眠是日常生活中重要的一环,现代医学研究成果表明睡眠中并不只是简单的"睡着了",宏观上睡眠可以划分为N-REM(非快速眼动期)和REM(快速眼动期)两阶段,两阶段呈现相互交替的周期分布。不同的睡眠状态具有截然不同的作用,并且睡眠状态的分布并不稳定,具有易被破坏的特点。睡眠周期不仅具有医学意义,在不同的睡眠阶段进行唤醒会对人产生显著的影响。现有的睡眠周期检测手段具有成本高、专业要求高、侵入性强的特点。能够进行轻量级、低成本、非侵入式的睡眠周期识别对改善用户的生活和对睡眠的认知都有显著的作用。随着智能手机搭载的传感器越来越多样化以及通信、计算能力的提升,越来越多的研究工作将目标设定在使用智能手机感知人们生活的方方面面。而睡眠由于用户意识下降的不可自知性和其重要的医学意义成为诸多研究的焦点。睡眠周期分布由于难以观察、识别线索弱、个体差异明显等不利因素,使得睡眠周期识别成为睡眠相关工作中的难点。本文对使用智能手机识别人的睡眠状态进行了探索。从医学文献中关于睡眠状态和用户夜间的外部表现研究成果及智能手机所搭载的传感设备所搭载的传感器的感知能力出发,将目标设定为日常生活场景中,对用户清醒以及用户睡眠期间的REM(快速眼动期)、N-REM(非快速眼动期)三种状态的识别。建立了完整的睡眠周期识别系统,从智能手机搭载的传感器以及系统运行状态获得数据,通过模式识别方法对用户睡眠期间发生的事件进行识别,结合医学背景中的知识提取这些事件的相应特征,通过序列化模式识别算法对用户的睡眠周期进行识别。并进行了大量实验构造了基于智能手机探测用户睡眠周期识别的数据集。本文的贡献在于:·实现了基于智能手机的睡眠周期识别系统:该系统通过智能手机搭载的传感系统对用户睡眠事件进行识别,在事件识别的基础上结合医学知识,采用决策融合的方式对用户的清醒、REM睡眠状态、N-REM睡眠状态进行识别。·进行了长期的真实场景下的实验:同时采用眼动仪监测和智能手机采集的方式,构建了使用智能手机识别睡眠周期的数据集。在实验数据集上对系统的有效性进行了验证,对实验中表现出的个体差异、模型适配性、模型随数据集规模的变化进行了分析。·基于睡眠周期识别系统,实现了基于睡眠周期识别结果改善用户生活的应用程序:智能唤醒闹钟和睡眠统计报告。
[Abstract]:Sleep is an important part of daily life. Modern medical research shows that sleep is not simply "asleep", but can be divided into N-REM (non-rapid eye movement) and REM (rapid eye movement) stages. The two phases present alternating periodic distributions. Different sleep states play a different role, and the distribution of sleep states is unstable and easy to be destroyed. Sleep cycle is not only of medical significance, but also has a significant effect on people during different sleep stages. The existing sleep cycle detection methods are characterized by high cost, high professional requirements and strong invasion. The ability to perform lightweight, low-cost, non-invasive sleep cycle identification plays a significant role in improving the user's life and awareness of sleep. With the increasing diversity of sensors, communication and computing power, more and more research has focused on the use of smart phones to perceive all aspects of people's lives. However, sleep has become the focus of many researches because of the decrease of user consciousness and its important medical significance. Because the distribution of sleep cycle is difficult to observe, the identification clue is weak, the individual difference is obvious and so on, the sleep cycle recognition becomes the difficulty in the sleep related work. This paper explores the use of smart phones to identify sleep states. Based on the research results of sleep state and night external performance of users in medical literature and the sensing ability of sensor devices carried on smart phones, the target is set in the daily life scene. Recognition of three states of REM (rapid eye movement) (REM) and non-rapid eye movement (NREM) during the user's sleep. A complete sleep cycle recognition system is established. The data are obtained from the sensors and the system running state on the smart phone, and the events occurring during the user's sleep are identified by the pattern recognition method. Combined with the knowledge in medical background, the corresponding features of these events are extracted, and the sleep cycle of users is identified by serialization pattern recognition algorithm. A large number of experiments were carried out to construct the sleep cycle identification data set based on smart phone detection. The contribution of this paper lies in the realization of the sleep cycle recognition system based on smart phone. The system recognizes the sleep events of the user through the sensor system of the smart phone, and combines the medical knowledge with the event recognition. Using the method of decision fusion to identify the sleep state of the user's awake REM sleep and N-REM sleep. The experiments were carried out under the long-term real scene: eye movement monitor and smart phone acquisition were used at the same time. A data set using smart phones to identify sleep cycles is constructed. The validity of the system is verified on the experimental data set, and the individual differences, model adaptability and the variation of the model with the size of the data set are analyzed, based on the sleep cycle recognition system. The application program of improving user's life based on the result of sleep cycle recognition is implemented: intelligent wake-up alarm clock and sleep statistics report.
【学位授予单位】：南京大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP212.9

【相似文献】