WKPowerMeans多径簇识别算法

针对KPowerMeans聚类算法多径散射簇的估计过程复杂及聚类结果高度依赖随机初始簇中心的问题,提出了一种改进的多径簇识别算法——WKPowerMeans算法.首先利用小波变换的尖峰检测技术估计出多径散射簇的数目和初始簇中心的位置,然后以结合了多径功率加权的多径分量距离为准则进行多径簇聚类.仿真结果表明:与KPowerMeans算法相比,采用所提出的WKPowerMeans算法能得到更稳定、准确的聚类结果,而且具有较低的时间复杂度.

WKPowerMeans Approach to Multipath Cluster Identification

In order to solve the problems of KPowerMeans multipath cluster recognition algorithm,which has a complex process of multipath scattering cluster estimation and whose clustering result is highly dependent on the ran-dom initial cluster cancroids. An improved algorithm,named WKPowerMeans,is proposed. The peak detection and information entropy methods are combined to develop the framework of automatic cluster identification. The im-proved algorithm not only acquires the number of cluster and initial centroids by using the wavelet transformation, but also adaptively obtains the different weights of the attributes of the multipath component. Simulation results indi-cate that the proposed WKPowerMeans clustering method can produce more robust and more accurate solutions than KPowerMeans method;furthermore it has lower time complexity.