一种基于K-L变换聚类的室内定位算法

摘要

When APs are intensively deployed in indoor location environment, the correlation of clustering center is high and clustering is not accurate owing to the high correlation of RSS received from neighboring APs.Aiming at solving the problem above, a clustering algorithm based K-L transform is put forward.The algorithm firstly treats the RSS fingerprint data with K-L transformation to remove the correlation and retain the greatest feature of it, and then k-means clustering algorithm is used to obtain higher clustering accuracy, thus to improve the positioning accuracy.The experimental results show that clustering accuracy is higher compared with the traditional clustering algorithm without K-L transforming.When determining that the dimension of RSS data after K-L transformation is 5 and the number of clustering center is 5 too, positioning error probability has increased by 9.3% within 2 m.%室内定位环境中接入节点(access point,AP)部署密集时,针对参考节点(reference point,RP)接收到邻近AP的接收信号强度(received signal strength,RSS)数据相关性大,而导致聚类过程中聚类中心相关性高、聚类不准确等问题,提出了一种基于K-L变换的聚类算法;该算法通过K-L变换对RSS指纹数据去相关处理来保留原始指纹数据最大的特征信息数据,然后通过k-means聚类算法聚类能够得到更高的聚类准确率,从而来提高定位精度.实验结果表明,该算法比没有经过K-L变换去相关处理的聚类算法聚类准确率要高;并且在实验过程中确定RSS数据经K-L变换降维之后的维数为5、聚类中心数为5时,定位误差在2 m以内的概率提高了9.3%.