正则化距离准则的Wi-Fi位置指纹室内定位方法

针对K-近邻法中常规指纹相似度匹配准则未能充分利用测试点和参考点的几何距离信息问题,该文提出了正则化距离准则的Wi-Fi位置指纹室内定位方法。该方法较常规距离准则兼顾了测试点与参考点的指纹距离和可靠几何距离,仅增加了正则化距离和K-近邻位置估算的迭代流程。而且,该方法仅含一个正则化因子,确定方法简便且可解释性好。密集与稀疏参考点格网间距下智能手机Wi-Fi平面定位实验表明,所提方法在正则化距离准则下能够有效提高约20%的定位精度,其中曼氏和欧氏定位中误差不超过2m。

A regularized-distance rule method for Wi-Fi fingerprinting indoor positioning

Aiming at the problem that traditional distance in K-Nearest Neighbor fingerprinting matching method does not use the geometry distance between test points and reference points,a regularizeddistance-rule method was proposed based on Wi-Fi fingerprinting indoor positioning.Comparing with traditional methods,the new method took both fingerprinting distance and credible geometry distance between test points and reference points into account,which only increased iterative processing of regularizing distance and estimated coordination using K-Nearest Neighbor.Moreover,the regularized-distance only contained one regularized factor which was clear and easy to understand.Finally,the indoor positioning experiment using smart phone under two grid scales of Wi-Fi fingerprinting was implemented at CUMT indoor and outdoor positioning experiment site.The result demonstrated that the positioning RMSE were generally decreased by about twenty percent using the proposed method including Manhattan,Chebyshev and Euclidean distance.Especially,positioning RMSE via Manhattan distance and Euclidean distance was below 2 meters on the case of dense grid.