基于深度残差适配网络的通信辐射源个体识别

Communication transmitter individual identification based on deep residual adaptation network

为解决通信辐射源识别中传统的人工特征提取方法鲁棒性不足和深度学习方法需要大量带标签目标域数据的问题,提出一种基于深度残差适配网络的通信辐射源个体识别方法。应用深度学习技术实现从源域到目标域上的迁移识别,只需要将带标签的源域数据和无标签的目标域数据进行训练。原始通信辐射源信号经过预处理后输入网络训练,将源域和目标域的分布差异和网络的损失函数作为优化目标,反复迭代得到最终模型。在实际采集的通信辐射源数据集上的实验结果证明了该方法的可行性和有效性。

In order to solve the problem that the traditional artificial feature extraction method is not robust enough and the deep learning method needs a large number of labeled target domain data,a communication transmitter individual identification method based on deep residual adaptation network is proposed.Applying deep learning technology to realize the transfer recognition from the source domain to the target domain only needs to train the labeled source domain data and the unlabeled target domain data.The original communication emitter signal is input into the network training after preprocessing.The distribution difference between the source domain and the target domain and the loss function of the network are taken as the optimization objectives,and the final model is obtained by iteration.The experimental results on the actual communication emitter data set show that the method is feasible and effective.