基于奇异值检测和AP聚类的室内指纹定位算法

摘要

To solve the problem caused by fluctuation of received signal strength indicator and large computing capacity in on‐line measurement ,an indoor localization algorithm based on singularity detection and affinity propagation clustering was proposed . Taking advantage of Hampel filter and KDE ,a method of singularity detection was proposed ,then affinity propagation clustering algorithm was used to cluster received signal strength indicator .Rough and meticulous localization method was used to locate the consumer .The algorithm proposed can improve localization accuracy and reduce computing capacity Comparing with the tradi‐tional method .%为克服基于无线局域网指纹定位算法中接收信号值波动和在线检测时计算量大的问题，提出一种基于奇异值检测和A P聚类的无线局域网指纹定位算法。分析 H am pel滤波器和核概率密度估计在接收信号强度值中奇异值检测的不足，结合Hampel滤波器和核概率密度估计两种方法在奇异值检测中的优势，给出一种奇异值检测算法；利用 AP聚类算法，对离线训练系统中的信号强度测量值进行聚类；通过AP聚类粗检测和基于加权 k近邻算法的细检测评估得到用户位置，完成定位。对比传统方法，该定位算法能够提高定位的准确性，降低算法的计算复杂度。