基于流形插值数据库构建的WLAN室内定位算法  

WLAN Indoor Localization Algorithm Based on Manifold Interpolation Database Construction

作　　者：周牧[1,2] 唐云霞[1] 田增山[1] 卫亚聪 

机构地区：[1]重庆邮电大学移动通信技术重庆市重点实验室,重庆400065 [2]天津师范大学天津市无线移动通信与无线电能传输重点实验室,天津300387

出　　处：《电子与信息学报》2017年第8期1826-1834,共9页Journal of Electronics & Information Technology

基　　金：国家自然科学基金(61301126);长江学者和创新团队发展计划(IRT1299);重庆市科委重点实验室专项经费;重庆邮电大学青年科学研究项目(A2013-31)~~

摘　　要：针对传统无线局域网(WLAN)室内定位系统中因参考点密集分布及逐点信号采集所带来的位置指纹数据库构建工作量繁重的问题,该文提出一种基于混合半监督流形学习和3次样条插值的数据库构建方法。该方法利用少量标记数据和大量未标记数据求解定位目标函数的最优解,同时根据高维信号强度空间与低维物理位置空间的映射关系,实现对未标记数据的位置标定。大量实验结果表明,该方法能够在保证较高定位精度的同时,显著降低位置指纹数据库的构建开销。To deal with the high cost involved in the location fingerprint database construction due to the dense Reference Points （RPs） distribution and point-by-point Received Signal Strength （RSS） collection in the conventional Wireless Local Area Network （WLAN） indoor localization systems, a new database construction approach based on the integrated semi-supervised manifold learning and cubic spline interpolation is proposed. The proposed approach utilizes a small amount of labeled data and a massive amount of unlabeled data to find the optimal solution to localization target function, and meanwhile relies on the mapping relations between the high-dimensional signal strength space and low-dimensional physical location space to calibrate the unlabeled data with location coordinates. The extensive experiments demonstrate that the proposed approach is able to guarantee the high localization accuracy, as well as significantly reduce the cost involved in location fingerprint database construction.

关 键 词：无线局域网 位置指纹 半监督学习 流形对齐 3次样条插值 

分 类 号：TN929.5[电子电信—通信与信息系统]