摘要
在辐射源个体识别应用中,受时间、空间、应用环境等因素的影响,辐射源个体特征会不断变化,不同时期得到的数据很难服从相同分布,而传统机器学习则要求数据服从相同分布。为了解决这一问题,提出一种基于迁移学习的辐射源个体识别分类方法。该方法通过聚类分析和重采样从数据集中选择新训练样本用于目标域学习,使用模糊近邻密度聚类提高对参数选择的鲁棒性及不同分布数据的适应性,并使用高斯核函数度量数据间的相似性以提高新训练样本选择的可靠性。实验结果表明,该方法能更有效、稳定地提高学习性能。
In the application of specific emitter identification,the specific emitter features will change for the influence of time,place and application environment,etc.,and it is difficult for data obtained in different periods to obey the same distribution,but traditional machine learning requires data to obey the same distribution.To solve this problem,a classification method based on transfer learning for specific emitter identification is proposed.In this method,new training samples are selected from the data set for target domain learning through cluster analysis and resampling,then fuzzy neighborhood density-based clustering is used to improve the robustness of parameter selection and the adaptability of different distribution data,and Gaussian kernel function is applied to measure the similarity among data to improve the reliability of new training samples.Experimental results demonstrate that the proposed method can more effectively and stably enhance the learning performance.
作者
刘振
陈阿磊
李世飞
袁俊泉
黄亮
LIU Zhen;CHEN A-lei;LI Shi-fei;YUAN Jun-quan;HUANG Liang(Air Force Early Warning Academy,Wuhan 430019,China)
出处
《舰船电子对抗》
2023年第2期53-60,78,共9页
Shipboard Electronic Countermeasure
关键词
辐射源个体识别
迁移学习
模糊近邻密度聚类
相似性度量
高斯核函数
specific emitter identification
transfer learning
fuzzy neighborhood density-based clustering
similarity measurement
Gaussian kernel function
