基于Wi-Fi的室内非法入侵检测识别算法研究

发布时间：2018-02-03 20:16

本文关键词： 室内安全入侵检测技术隐马尔科夫模型　出处：《太原理工大学》2016年硕士论文　论文类型：学位论文

【摘要】：室内安全关系到每个人的生命财产,一直是大众关注的焦点问题。学者们针对此问题提出了多种解决方案,常见的有视频监控和红外线监控。然而此类方案均需要额外的特制设备,如视频监控需要在室内中布置摄像头,而红外线监控需要布置红外线发生器。类似的额外设备开销限制了其应用推广,因而亟需一种开销小且可靠性高的解决方案。随着智能设备的普及,技术日渐受到大众的青睐。从商场、办公室到家庭,随处可见的踪影。如此普及的应用基础给学者们提供了广阔的想象空间,近年来各种基于的算法和系统层出不穷,如手势识别、室内定位和睡眠监测等。已经有学者关注人类检测问题并提出了相应的检测算法,然而这类算法只能判断是否有人进入某一感兴趣的区域,无法自动判断识别出进入者是用户还是入侵者,达不到理想结果。基于这一目标,本文提出一种基于的非法入侵检测识别算法,可以实时可靠地检测识别入侵者。理论依据为不同的用户有不同的生活习惯,具体表现为何时、何地做了何种动作。时间可以通过时间戳获得,地点可以通过总结信号随距离的衰减规律获得,用户动作会引起信号波动,相同动作引起的信号波动程度相似,因此动作可以通过学习训练信号波动逆向推导获得。至此,就成功为不同用户“量身定制”行为习惯模型。整个算法分为四个模块,分别是信号预处理模块、模型建立模块、入侵检测模块和报警确认模块。信号预处理模块的作用是先使用主成分分析法对信号去噪,再使用滑动窗口自下而上算法对信号分段,成为不同的信号片段序列;模型建立模块基于不同用户有不同生活行为习惯这一前提,使用隐马尔科夫模型为不同用户建模;入侵检测模块使用阈值检测识别非法入侵者;报警确认模块通过识别预定义手势,判断是否为误识别,从而进行报警确认。最后通过实验验证算法的有效性,识别精度最高可达,能够最大限度地减少用户财物的损失,达到了预期的效果。
[Abstract]:Indoor safety concerns everyone's life and property, which has always been the focus of public attention. Scholars have put forward a variety of solutions to this problem. Video surveillance and infrared surveillance are common. However, such schemes require additional specialized equipment, such as video surveillance, which requires the installation of cameras indoors. Infrared monitoring requires the arrangement of infrared generators. Similar additional equipment costs limit its application, so it is urgent to have a low cost and high reliability solution. With the popularity of intelligent devices. Technology is increasingly popular with the public. From shopping malls, offices to homes, everywhere can be seen. Such a popular application of the basis for scholars to provide a broad space for imagination. In recent years, a variety of algorithms and systems based on the emerging, such as gesture recognition, indoor positioning and sleep monitoring. However, this kind of algorithm can only judge whether someone enters a certain area of interest, and can not automatically judge whether the entrant is a user or an intruder, and can not achieve the desired result. In this paper, we propose an algorithm based on illegal intrusion detection, which can detect and identify intruders in real time and reliably. The theoretical basis is that different users have different living habits, and the specific performance is when. Where to do what action. Time can be obtained through the time stamp, the location can be summed up by the attenuation of the signal with the distance to obtain, the user action will cause signal fluctuations, the same action caused by the same signal fluctuation degree. So the action can be obtained by learning the inverse derivation of training signal fluctuation. So far, the behavior habit model can be "customized" for different users successfully. The whole algorithm is divided into four modules, which are signal preprocessing module. The model building module, intrusion detection module and alarm confirmation module. The function of the signal preprocessing module is to use principal component analysis to de-noise the signal first, and then to segment the signal by sliding window bottom-up algorithm. Become different signal fragment sequences; The model building module is based on the premise that different users have different living habits and uses hidden Markov model to model different users. Intrusion detection module uses threshold detection to identify illegal intruders; The alarm and confirmation module recognizes the predefined gesture and determines whether it is a false recognition. Finally, the validity of the algorithm is verified by experiments, and the recognition accuracy is the highest. It can minimize the loss of user property and achieve the desired results.
【学位授予单位】：太原理工大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TN92

【相似文献】