基于多域特征融合学习的辐射源个体识别算法  

Specific Emitter Identification Based on Multi-Domain Feature Fusion Learning

作　　者：寸陈韬 李天昀[1] 朱家威 查雄 CUN Chentao;LI Tianyun;ZHU Jiawei;ZHA Xiong(Information Engineering University,Zhengzhou 450001,China)

机构地区：[1]信息工程大学,河南郑州450001

出　　处：《信息工程大学学报》2022年第3期277-285,共9页Journal of Information Engineering University

摘　　要：现有的基于专家经验的特定辐射源个体识别(Specific Emitter Identification,SEI)方法和基于深度学习的SEI方法,通常在单一类型辐射源畸变存在的场景下性能较好,然而在多种辐射源畸变同时存在的复杂场景下表现较差。为此,提出一种基于多域特征融合学习的辐射源个体识别算法,将原始接收辐射源信号转换为I/Q眼图、矢量图和Hilbert-VMD时频谱图等多域信号表示作为网络输入,并结合神经网络进行多域特征融合提取。实验结果表明,与现有的基于专家经验的SEI算法或其他单一信号表示输入的基于深度学习的SEI算法相比,该算法在符号信噪比10 dB下的识别增益约10%。Existing specific emitter identification(SEI)methods based on expert experience and SEI methods based on deep learning generally perform better in scenarios where a single type of emitter distortion is present,but less well in complex scenarios where multiple emitter distortions are present simultaneously.Based on the above issues,this paper proposes a multi-domain feature fusion learning-based algorithm for SEI,which converts the received emitter signals into several signal representations in different domains such as I/Q eye diagram,vector diagram and HilbertVMD time-frequency spectrogram as the net input,and combines deep neural networks for multi-domain feature fusion extraction.Experimental results show that the algorithm is superior to the traditional SEI algorithm based on expert experience or other similar deep learning-based SEI algorithms with a single network input,with a recognition gain of approximately 10%at a symbolic signal-tonoise ratio of 10 dB.

关 键 词：辐射源个体识别 信号表示 深度学习 特征融合 

分 类 号：TN911.7[电子电信—通信与信息系统]