针对无线信道动态衰落特性引起的蜂窝网室内定位误差较大的问题,该文提出基于密度的空间聚类(Density Based Spatial Clustering of Applications with Noise,DBSCAN)子空间匹配算法,有效剔除大误差点,提高定位精度。首先通过划分信号空间,构建多个子空间,在子空间中利用加权K近邻匹配算法(Weighted K Nearest Neighbor,WKNN)估计出目标位置;然后利用DBSCAN对估计位置进行聚类以剔除异常点;最后结合概率模型确定最终估计位置。实验结果表明,基于DBSCAN的子空间匹配算法能有效剔除大误差点,提高蜂窝网室内定位系统的整体性能。
For the sake of reducing the indoor localization errors caused by dynamic signal fading in cellular network, this paper propose a novel Density-Based Spatial Clustering of Applications with Noise (DBSCAN) based subspace matching algorithm for indoor localization, which can improve the localization accuracy by eliminating the location with large errors. Specifically, the signal space is firstly divided into several subspaces, where a position estimation can be obtained respectively using the Weighted K Nearest Neighbors (WKNN) approach. Then, DBSCAN is applied to the position coordinates obtained from each subspace which eliminates the outliers. Finally, the location is estimated based on probability analysis. Experimental results show that the proposed approach can improve the location accuracy by eliminating the location with large errors.