【作者】 徐强；

【导师】 吴哲夫；

【作者基本信息】 浙江工业大学 ， 信息与通信工程， 2017， 硕士

【摘要】 室内无源定位在安防、智能家居等领域都具有一定的应用价值。作为主流的物理信号,接收信号强度存在精度低、不够稳定等缺点,而信道状态信息是物理层的信号,具有比接收信号强度更高的精度和更好的稳定性。本文对信道状态信息的原理和特点进行了分析,并设计了基于信道状态信息的无源定位方法。为进一步提高系统性能,本文提出了提高系统总体分类准确率的置信度方法。实验结果表明,所提出的定位方法在不同环境下都可获得较好的性能。另外,为了更好的感知室内人体状态,本文还对人体的朝向检测进行了研究,为后续工作打下了基础。本文的研究工作主要包括:(1)对传统的基于接收信号强度的定位方法进行了介绍,引入了信道状态信息并对其进行了讨论,分析了其在精度和稳定性方面的优势以及应用于无源室内定位的原理;(2)研究了信道状态信息数据的预处理和特征值提取等方法,使得采集到的数据更加可靠和有效;(3)根据数据特点及无源定位系统的要求,权衡比较了各种分类算法的优缺点,选定了朴素贝叶斯方法作为无源定位的算法;(4)提出了置信度方法以进一步提高室内人体位置的总体分类准确率,通过多个不同场景下的实验验证表明该方法最高可提升30%左右的性能,并且其在不同的环境下都具有较好的适应性;(5)提出了采用信道状态信息实现人体朝向的无源检测方案,通过实验证明朝向检测的可行性,四个方向最优的检测准确率在90%以上,并且分析了若干指标对检测准确率的影响。更多还原

【Abstract】 Passive indoor localization plays an important role in applications such as security and smart housing.Received Signal Strength Indicator is the most popular physical parameter utilized in indoor localization.However,Received Signal Strength Indicator is coarse-grained and not stable enough.Instead,Channel State Information outperforms Received Signal Strength Indicator with fine granularity and high stability.In this paper,we first explore and discuss the principle and characters of Channel State Information,and then design methods to realize passive indoor localization.To enhance the performance of the system,we propose level of confidence.As the results of experiments reveal,our schemes can achieve a relatively excellent performance in different indoor environments.What’s more,to obtain more detailed feature of human’s activity,we study orientation detection,which lays a foundation for our future work.The contribution for this paper mainly consists of the following parts:(1)We introduce some traditional schemes of indoor localization based on Received Signal Strength Indicator.Then Channel State Information is proposed.We discuss the superiority of Channel State Information from its fine granularity and stability as well as the reasons why it can be used for passive localization;(2)To make the data more reliable and efficient,we study some methods to pre-process and extract features from the original data;(3)According to the characters of the data and the demand of localization system,we compare and balance several classification algorithms,and finally select Naive Bayes as our basic algorithm during the phase of passive localization;(4)We propose level of confidence by combining the results from different antenna pairs.It can help enhance the total accuracy of localization system with an improvement of about 30% at best case.The experiments from several scenarios also demonstrate that our schemes can adapt well in diverse indoor environments;(5)We realize passive detection of human’s orientation.Experiments are carried out to prove feasibility and some factors which may affect the accuracy of detection are analyzed.The accuracy for 4 orientations detection at best case is more than 90%.更多还原