基于独立成分分析与核典型相关分析的WLAN室内定位方法  

WLAN indoor positioning algorithms based on independent component analysis and kernel canonical correlation analysis

作　　者：张勇[1,2] 史雅楠[1] 黄杰[1] 李飞腾[1] Zhang Yong Shi Yanan Huang Jie Li Feiteng(School of Computer ＆ Information, Hefei University of Technology, Hefei 230009, China Post-Doctoral Research Center of Wuhu Overseas Student Pioneer Park, Wuhu Anhui 241000, China)

机构地区：[1]合肥工业大学计算机与信息学院,合肥230009 [2]芜湖创业园留学人员博士后科研工作站,安徽芜湖241000

出　　处：《计算机应用研究》2016年第12期3817-3821,共5页Application Research of Computers

基　　金：国家科技支撑计划资助项目(2013BAH52F01)

摘　　要：接收信号强度(received signal strength,RSS)在WLAN室内定位环境中存在时变特性,降低了WLAN定位环境中RSS信号与位置信息之间的相关性,致使定位精度降低。针对这一问题,提出通过利用独立成分分析(independent component analysis,ICA)对RSS信号进行数据降维和去相关处理,提取独立分量;然后采用核典型相关分析(kernel canonical correlation analysis,KCCA)来提取独立分量与位置信息之间的典型相关特征;最后结合传统定位算法如加权K近邻法(weighted K nearest neighbors,WKNN)、支持向量机(support vector machine,SVM)算法实现定位。实验结果表明,传统定位算法WKNN、SVM算法通过运用ICA与KCCA特征提取后再进行定位其定位精度得到了提高。The time-varying character of received signal strength （RSS） in WLAN positioning environment drastically reduces the relevance between RSS signal and position information which result in the low positioning accuracy. Based on this situation, this paper used independent component analysis （ICA） algorithm to reduce the dimensions of RSS signal and their correlation, and extracted the independent components; then applied kernel canonical correlation analysis （KCCA） to extract the most correlated canonical features between the independent components of RSS signal and position information; finally employed the traditional positioning algorithms such as weighted K nearest neighbors （WKNN）, support vector machine （SVM） to localization. The experimental results show that with the deployment of ICA and KCCA to extract correlated canonical features, traditional WKNN and SVM positioning algorithms achieve better localization accuracy.

关 键 词：无线局域网 室内定位 接收信号强度 独立成分分析 核典型相关分析 

分 类 号：TN915.05[电子电信—通信与信息系统]