基于模糊核c-means算法的位置指纹聚类  

Location fingerprint clustering based on fuzzy kernel c-means algorithm

作　　者：李方[1] 佟为明[1] 李凤阁[1] 王铁成[1] 

机构地区：[1]哈尔滨工业大学电气工程及自动化学院,哈尔滨150001

出　　处：《控制与决策》2012年第8期1180-1184,1190,共6页Control and Decision

基　　金：国家自然科学基金项目(51077015);黑龙江省自然科学基金项目(E200914)

摘　　要：提出一种针对位置指纹的模糊核c-means聚类算法.将位置指纹归结为一种服从正态分布的区间值数据以反映接入点信号强度采样值的不确定性,通过区间中值和大小确定的正态分布函数将位置指纹映射为特征空间中的一点,并在该特征空间中采用基于核方法的模糊c-means算法对其进行聚类.通过ZigBee定位实验表明,该方法对于位置指纹的分类效果明显好于基于信号强度平均值的c-means聚类,可在保证定位精度的前提下有效降低定位的计算量.A fuzzy kernel c-means clustering algorithm（FKC） is proposed to resolve the location fingerprint（LF） clustering. LF is summarized as a kind of interval-valued data which obey normal distribution to describe sampling uncertainty of received signal strength of access point. After mapping LF into the high-dimensional feature space through normal distribution function determined by interval median and size, LF is clustered with fuzzy c-means algorithm based on kernel method in the feature space. Results of ZigBee positioning experiments show that FKC can get better clustering effect than c-means algorithm based on the average value of signal strength. On the premise of ensuring the positioning precision, a feasible solution is provided to decrease the positioning calculation consumption remarkably.

关 键 词：位置指纹聚类 区间值数据 核方法 模糊c-means 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]