基于深度学习的辐射源个体识别方法综述

Overview of Specific Emitter Identification Methods Based on Deep Learning

首先,从辐射源的个体识别过程的3个步骤,即信号预处理、信号特征提取、信号分类分别展开分析,阐述了将深度学习应用于辐射源的个体识别方法;然后,归纳了基于深度学习的辐射源识别方法的改进趋势,包括零小样本学习的弱监督学习方法和多特征融合识别;最后,针对神经网络的训练和推理均需要消耗大量算力的问题,总结了基于现场可编程门阵列(FPGA)处理器的深度学习加速器并将其运用于本识别方法中,从而提高辐射源个体识别的速率、降低功耗和硬件成本,使基于深度学习的辐射源个体识别方法,能以更好的性能应用于实际场景中。

Firstly,from three steps of the individual identification process of emitters,namely signal preprocessing,signal feature extraction and signal classification,the specific emitter identification method applying deep learning are expounded.Then,the improvement trend of emitter identification method based on deep learning is summarized,which includes weakly supervised learning method with zero-small sample learning and multi-feature fusion recognition.Finally,for the problem that the training and inference of neural networks need to consume a lot of computing power,the deep learning accelerator based on field programmable gate array(FPGA)processor is summarized and applied to the identification method,thereby the rate of specific emitter identification is improved,as well as the power consumption and hardware costs are reduced,so that the specific emitters identification method based on deep learning can be applied to actual scenarios with better performance.