基于运动传感器的日常生活智能监护

发布时间：2018-06-09 06:35

本文选题：智能家居 + 无线传感器网络　；参考：《重庆大学》2014年硕士论文

【摘要】：目前我国正处于老龄化社会阶段，由于子女大多出外工作，老人家庭空巢率也在不断增加，对智能化的看护系统的需求更加紧迫。而智能化看护系统的关键问题就在于对老人在日常生活中的活动进行识别和理解。目前，针对活动识别的研究主要可分为两大类：基于视觉监控设备的方式和基于传感器设备的方式。基于视觉监控设备的方式虽然在实现技术上已经比较成熟，但由于其在采集数据的过程中侵犯了观察对象的隐私，因此并不适合在家居环境中采用。而随着无线传感网络的发展，，利用无线传感设备收集活动数据，进行活动识别，已经吸引了越来越多的研究者的注意力。本文将活动识别和无线传感器网络技术结合起来，针对家居环境中日常生活的智能监护问题，提出了异常活动分布式检测方法DetectingAct。由于正常活动与异常活动的区分是一个比较主观的问题，本文中的正常活动被定义为在活动数据中反复出现的活动，而异常活动被定义为在时空数据上与正常活动存在着较大偏差的活动。本文的主要贡献体现在以下几个方面： ①针对日常生活智能监护中对实时性的要求，本文设计了分布式异常活动检测方法DetectingAct。该方法在检测时充分利用了传感器节点自身有限的计算资源和存储资源，避免了集中式检测方法带宽需求大和反应时间长的缺点，同时保证了检测精度，提高了检测速度。 ②针对传统活动识别中对活动定义的缺陷，本文在活动模型的原有轨迹信息的基础上，引入了触发数据中的持续时间信息。改进后的模型对活动的定义更准确，提高了检测精度。 ③针对当前研究中缺乏面向运动传感器的智能环境仿真系统的问题，开发设计了日常生活仿真及统计系统作为实验基础平台。该系统可对真实环境下的智能环境进行仿真，所获取到的仿真数据的可信度较高。 ④利用从日常生活仿真及统计系统中生成的仿真数据与真实环境中产生的触发数据进行了实验，从准确性、实时性、稳定性三个方面验证了分布式异常活动检测方法DetectingAct相比于传统的基于轨迹的检测算法的优势。
[Abstract]:At present, our country is in the stage of aging society, because most of the children go out to work, the empty nest rate of the elderly family is also increasing, so the need for the intelligent nursing system is more urgent. The key problem of intelligent nursing system is to identify and understand the activities of the elderly in their daily life. At present, the research on activity recognition can be divided into two main categories: visual monitoring devices and sensor devices. Although the method based on visual monitoring device is mature in technology, it is not suitable for home environment because it infringes the privacy of observation object in the process of collecting data. With the development of wireless sensor network (WSN), it has attracted more and more researchers' attention to collect activity data and identify activities by wireless sensor devices. In order to solve the problem of intelligent monitoring of daily life in home environment, a distributed detection method for abnormal activities is proposed. Since the distinction between normal and abnormal activities is a more subjective problem, the normal activities in this paper are defined as recurring activities in the activity data. The abnormal activity is defined as the activity which deviates greatly from the normal activity in time and space data. The main contributions of this paper are as follows: 1 according to the requirement of real time in intelligence monitoring of daily life, In this paper, a distributed anomaly detection method, detect activity, is designed. This method makes full use of the limited computing and storage resources of sensor nodes in detection, avoids the shortcomings of large bandwidth and long reaction time of centralized detection methods, and ensures the accuracy of detection. The detection speed is improved. 2 aiming at the defect of the definition of activity in traditional activity recognition, this paper introduces the duration information of trigger data on the basis of the original trajectory information of the activity model. The improved model is more accurate in the definition of activity and improves the accuracy of detection. 3 aiming at the lack of intelligent environment simulation system for motion sensor in current research, The daily life simulation and statistics system is developed as the experimental platform. The system can simulate the intelligent environment in real environment. The credibility of the obtained simulation data is high. 4 the simulation data generated from the daily life simulation and statistical system and the trigger data generated in the real environment are used for experiments. Three aspects of stability verify the advantages of the distributed anomaly detection method (detection Act) over the traditional locus based detection algorithm.
【学位授予单位】：重庆大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP277;TP212.9;TN929.5

【参考文献】