数据库样本缺失下的雷达辐射源识别

Radar Emitter Identification in Database Sample Missing Condition

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。

Present radar emitter identification based on machine learning technology mostly assumes that training set and test set are same. When the radar database and the true distribution of the signals are bi- ased,the traditional classification method is ineffective. Thus, the theory of transfer learning is introduced into the identification system,and a radar emitter signal identification method based on structural discovery and re-balancing is proposed. By means of database data and target data clustering analysis and resam-pling ,the distribution is corrected and the new data is put to support vector machine （ SVM） for training and identifying reconnaissance samples. The simulation results show that the classification performance of the support vector machine model in the new training sample set has been greatly improved.