摘要
由于通信电台信号的样本小,电台指纹特征弱,导致通信电台的个体识别准确度不高,本文首次提出了基于密度峰值算法进行通信电台个体识别,在不需要训练样本的条件下就能对通信电台进行个体识别。首先对信号进行矩形积分双谱变换,提取信号1×L维矩形双谱特征,计算各个信号间的欧式距离,然后根据密度峰值算法的定义计算各个信号的密度ρ和δ,以ρ、δ为横坐标与纵坐标画二维图,找到聚类中心,对各个信号进行分类识别。与传统的通信电台分类识别方法相比,此方法运用的是机器学习中聚类的方法,是无监督的方法,不需要带标签的通信电台信号样本,在实际运用中会发挥更大的作用。
Due to the small sample of communication station signals and the weak fingerprint characteristics of radio stations,the accuracy of individual identification of communication stations is not high.This paper firstly proposes the identification of communication stations based on density peak algorithm,which can be used without training samples.Firstly,the signal samples are subjected to rectangular integral bispectral transformation,and the 1×L-dimensional rectangular bispectrum features are extracted.Then the Euclidean distance between each signal is calculated,and the densityρandδof each signal are calculated according to the definition of the density peak algorithm.Withρandδas a two-dimensional map is drawn on the abscissa and the ordinate,and the cluster center is found,and each signal is classified and identified.Compared with the traditional communication station classification and recognition method,this method uses the clustering method in machine learning,which is an unsupervised way.It does not need samples signal from a communication station with a label,and it will play a greater role in practical application.
作者
李昕
雷迎科
Li Xin;Lei Yingke(Electronic Countermeasures Institution of National University of Defense Technology,Hefei,Anhui 230037,China)
出处
《信号处理》
CSCD
北大核心
2019年第7期1242-1249,共8页
Journal of Signal Processing
关键词
矩形积分双谱变换
密度峰值算法
机器学习聚类
不需要电台信号样本
rectangular integral bispectral transform
density peak algorithm
machine learning clustering
no need for radio signal sample
