基于KFCM-LMC-LSSVM算法的WLAN室内定位方法  

Indoor Positioning Algorithm for WLAN Based on KFCM-LMC-LSSVM

作　　者：王昱洁[1] 王媛 张勇[1] WANG Yu-jie;WANG Yuan;ZHANG Yong(School of Computer and Information, He Fei University of Technology, Hefei, Anhui 230009, Chin)

机构地区：[1]合肥工业大学计算机与信息学院,安徽合肥230009

出　　处：《计量学报》2018年第4期554-558,共5页Acta Metrologica Sinica

基　　金：国家科技支撑计划(2013BAH52F01)

摘　　要：针对WLAN室内定位采集指纹点工作量大且定位精度不高的问题,提出一种基于核模糊C均值聚类(kernelized fuzzy C-means,KFCM)、低秩矩阵填充(low-rank matrix completion,LMC)及最小二乘支持向量机(least squares support vector machine,LSSVM)的室内定位算法。首先将指纹点利用KFCM算法进行聚类,使待测点定位到一个区域内。在该区域里运用LMC理论,重构出具有高密度指纹点的指纹库。最后利用LSSVM定位出待测点的物理位置。实验结果表明,采用KFCM-LMC-LSSVM算法不仅减少了构建指纹库的工作量,而且提高了定位精度。To solve the problems of large workload whe the fingerprint database is constructed and low positioning accuracy in WLAN indoor positioning system. An indoor positioning algorithm was presented, which integrates kernelized fuzzy C-means（ KFCM）, low rank matrix completion （LMC） and least squares support vector machine（LSSVM）. Firstly, the KFCM is used to cluster the fingerprint points, the test points are sorted into one of the small areas. According to the LMC theory, the low rank fingerprint database into high density fingerprint database was translated. Finally the position of the test points is determined by the LSSVM algorithm. Experiments showed that the KFCM-LMC-LSSVM algorithm not only reduces the workload of building the fingerprint library, but also has higher positioning accuracy.

关 键 词：计量学 室内定位精度 核模糊C均值聚类 低秩矩阵填充 不精确拉格朗日乘子法 

分 类 号：TB973[一般工业技术—计量学]