数据库样本缺失下的雷达辐射源识别

发布时间：2018-05-15 08:21

本文选题：雷达辐射源识别 + 迁移学习　；参考：《电讯技术》2017年07期

【摘要】：目前,基于机器学习的雷达辐射源识别技术大多以训练集和测试集同分布为假设,当雷达数据库样本不足导致与信号真实分布存在偏差时,传统的分类方法效果不佳。为此,将迁移学习理论引入识别系统,设计了一种基于结构发现与再平衡的雷达辐射源信号识别方法。通过对数据库和待识别辐射源信号样本进行聚类分析发现数据结构信息,通过重采样处理修正其分布差异。将新采样数据输入支持向量机进行训练并对侦收样本进行识别。仿真实验表明,在新训练样本集上学习的模型对测试集的分类性能有了很大的提升。
[Abstract]:At present, most radar emitter recognition techniques based on machine learning assume the same distribution of training set and test set. When the lack of radar database samples leads to deviation from the real signal distribution, the traditional classification method is not effective. Therefore, the migration learning theory is introduced into the recognition system, and a radar emitter signal recognition method based on structure discovery and rebalance is designed. The data structure information is found by clustering analysis of the database and the samples of emitter signals to be identified, and the distribution difference is corrected by resampling processing. The new sampling data is input into the support vector machine to train and recognize the collected samples. The simulation results show that the learning model on the new training sample set can greatly improve the classification performance of the test set.
【作者单位】：装备学院研究生管理大队;装备学院光电装备系;
【分类号】：TN957.51

【相似文献】