摘要
日益复杂的电磁环境对通信辐射源个体识别提出了更高挑战,而传统的特征提取方法难以满足需求。深度学习因其在特征表示方面的优势,在多个领域的识别任务中获得了巨大成功。针对辐射源个体识别任务,利用深度神经网络,结合自然语言处理与计算机视觉领域的注意力机制思想,将双注意力机制引入预处理层与特征提取层实现对神经网络的优化,并针对同型号电台进行个体识别。实验结果表明,所提方法的识别效果比基于传统特征的方法和普通深度神经网络方法有较明显的提升。
The increasingly complex electromagnetic environment poses a higher challenge to the individual identification of communication emitter,but the traditional feature extraction methods are difficult to meet the requirements.Due to its advantages in feature representation,deep learning has achieved great success in many fields.Aiming at the task of emitter individual recognition,by using deep neural network,combined with the idea of attention mechanism in the field of natural language processing and computer vision,this paper introduces dual attention mechanism into the preprocessing layer and feature extraction layer to optimize the neural network,and carries out individual recognition for the same type of radio station.The experimental results indicate that the recognition effect of the proposed method is significantly improved compared to the traditional feature-based method and the ordinary deep neural network method.
作者
张宸嘉
朱磊
陈璞
俞璐
ZHANG Chenjia;ZHU Lei;CHEN Pu;YU Lu(Army Engineering University of PLA,Nanjing Jiangsu 210007,China)
出处
《通信技术》
2021年第7期1594-1600,共7页
Communications Technology
关键词
深度神经网络
注意力机制
辐射源个体识别
计算机视觉
自然语言处理
deep neural network
attention mechanism
emitter individual recognition
computer vision
natural language processing