一种基于改进K均值算法的跳频信号分选方法

摘要

To avoid converging to local minima while determing the number of clusters, this paper proposed a modified initial cluster centers selection algorithm.This algorithm firstly searched the peaks of statistic histogram by morphological processing,and then correctly estimated the number and the initial location of cluster centers according to a merge threshold.Frequency hopping signal could be sorted by combining the proposed selection algorithm with conditional K-mean algorithm.Experimental results show that compared with other K-mean algorithm, the proposed modified K-mean algorithm is able to sort FH signals with high correctness.%为了在估计聚类数目的同时避免收敛到局部极小值,提出了一种改进的初始聚类中心选取算法.该算法通过形态学处理搜索统计直方图的峰值,根据合并门限正确估计聚类中心的数目和初始位置.将其与传统的K均值算法相结合,可用于跳频信号分选.实验结果表明,与其他K均值算法相比,该改进K均值算法能够以很高的正确率分选跳频信号.