基于多维尺度分析的自适应室内群终端定位方法  

Adaptive Indoor Location Method for Multiple Terminals Based on Multidimensional Scaling

作　　者：付先凯 蒋鑫龙[1,2,3] 刘军发 张少博[4] 陈益强 FU Xian-kai;JIANG Xin-long;LIU Jun-fa;ZHANG Shao-bo;CHEN Yi-qiang(Center of Pervasive Computing System Research,Institute of Computing Technology,Chinese Academy of Sciences,Beijing 100190,China;Beijing Key Laboratory of Mobile Computing and Pervasive Device,Chinese Academy of Sciences,Beijing 100190,China;University of Chinese Academy of Sciences,Beijing 100049,China;School of Information Engineering,Chang’an University,Xi’an 710064,China)

机构地区：[1]中国科学院计算技术研究所泛在计算系统研究中心,北京100190 [2]中国科学院移动计算与新型终端北京市重点实验室,北京100190 [3]中国科学院大学,北京100049 [4]长安大学信息工程学院,西安710064

出　　处：《计算机科学》2018年第10期104-110,共7页Computer Science

基　　金：国家自然科学基金面上项目:面向可穿戴用户行为识别的增量学习方法研究(61572471);融合多元传播模型和指纹模型的免标定室内定位方法研究(61472399);递归深度学习网络的多极限环神经动力学模型及方法研究(61572004);广东省科技计划项目:面向健康监护的新型智能贴件关键技术研发及产业化(2015B010105001)资助

摘　　要：室内定位是普适计算领域的热点研究问题。当前,室内定位方法主要分为基于信号传播模型的定位方法和基于无线信号指纹的定位方法。其中,基于指纹的方法由于不需要知道无线信号接入点(Access Point,AP)的位置而得到更加广泛的应用,其需要通过离线阶段采集大量数据来构建丰富的指纹库,满足这一条件需要大量的人工标定工作。对此,文中提出了一种基于指纹空间关系的定位方法,相比于传统的指纹定位方法,该方法无需建立指纹库,只需要通过获取多终端的Wi-Fi信号强度,计算所有终端的不相似度并构建不相似矩阵;通过多维尺度分析(Multidimensional Scaling,MDS)算法,构建出所有终端的位置分布图,进而通过确定其中3个以上终端的位置来定位所有的终端。采用支持向量回归机(Support Vector Regression,SVR)计算任意终端间的距离,并将距离矩阵作为不相似矩阵。文中在商场场景下选择了约2500m^2的区域进行实验,所提方法的平均定位误差约为7m。Indoor location is a hot research topic in the field of pervasive computing.At present,indoor location methods are mainly divided into the localization method based on signal propagation model and the one based on wireless signal fingerprint.The fingerprint based method is more widely used because it does not need to know the location of the wireless signal AP.But it needs to collect a large amount of data at the offline stage to build a rich fingerprint database,which needs a lot of manual calibration.For this reason,this paper proposed a localization method based on spatial relations of fingerprints.Compared with the traditional fingerprint localization methods,this method does not need to build a fingerprint database.It uses Wi-Fi fingerprint from multiple terminals to extract the similarity of fingerprints and construct a dissimilarity matrix,and finally applies multidimensional scaling(MDS)algorithm to construct the relative location map for all terminals.Then each terminal can be positioned by determining the position of more than 3 terminals.In this paper,support vector regression(SVR)is used to calculate the distance between arbitrary terminals,and the distance matrix is used as the dissimilarity matrix.A shopping mall which is about 2500 square meter is selected as testing environment,and the average positioning error of the proposed method is about 7 meters.

关 键 词：室内定位 指纹定位方法 多维尺度分析(MDS) 支持向量回归机(SVR) 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]