自适应动态K的WKNN室内定位方法  

Self-adaptive WKNN indoor location method

作　　者：胡久松[1] 刘宏立[1] 徐琨[1] HU Jiusong;LIU Hongli;XU Kun(College of Electrical and Information Engineering,Hunan University,Changsha 410082,P.R.China)

机构地区：[1]湖南大学电气与信息工程学院,长沙410082

出　　处：《重庆邮电大学学报（自然科学版）》2018年第4期431-438,共8页Journal of Chongqing University of Posts and Telecommunications(Natural Science Edition)

基　　金：中央国有资本经营预算项目(财企[2013]470号);国家自然科学基金(61771191);中央高校基本科研项目(1053214004);湖南省自然科学基金项目(2017JJ2052);教育部产学合作协同育人项目(201601004010;201701056026);湖南省普通高校教学改革研究项目(湘教通[2016]400号);湖南省研究生创新项目(CX2017B112)~~

摘　　要：WiFi室内定位已被广泛研究,并且提出了许多解决方案,其中以接收信号强度(received signal strength,RSS)作为位置指纹的加权K-最近邻(weighted K-Nearest neighbor,WKNN)算法是目前使用最广泛的位置指纹算法之一。由于WKNN算法通常采用固定的K值,其定位精度在实际使用时具有局限性。尽管动态K的方案被提出,但是由于引入了新的不确定性参数,因此,并未真正解决问题。针对这个问题,提出了一种自适应动态K的WKNN室内定位方法。提出的算法的K值自适应调整仅依赖于离线和在线数据,即可以不引入新的不确定参数。在这个前提下,提出的算法采用"多雷达搜索策略"的方式自适应选择近邻数K值进行在线位置估计。在真实环境中采样了大量数据进行了试验。试验结果表明,提出的算法可根据在线情况自适应调整K值,获得了较好的定位结果。WiFi indoor positioning has been extensively studied and many solutions have been proposed.The weighted K-nearest neighbor(WKNN)algorithm with received signal strength(RSS)as the position fingerprint was one of the most commonly used fingerprint algorithms.As the WKNN algorithm usually uses a fixed K value,its positioning accuracy has been limited when it is practically applied.Although the dynamic K scheme was proposed,the problem was not really solved due to the introduction of new uncertain parameters.Aiming at solving this problem,a self-Adaptive WKNN indoor locating method is proposed.The proposed algorithm adjusts K values only depends on the offline and online data,i.e.it will not introduce new uncertain parameters.In this context,the algorithm uses the“multi-radar search strategy”to adapt to the selection of K value of the nearest neighbor to estimate the position.The experimental results showed that the proposed algorithm could adjust the K value adaptively according to the online condition,and obtain a fine positioning accuracy.

关 键 词：室内定位 指纹识别 加权K-最近邻(WKNN) 无线电地图 相似度度量 

分 类 号：TN393.17[电子电信—物理电子学]