一种新的基于信号导频的射频指纹识别方法

现有基于深度学习的射频指纹识别技术大多采用原始数据样本作为网络输入，未考虑信号携带内容对分类结果产生的影响，网络结构相对单一。为此，针将信号导频部分作为网络输入展开了研究，提出了一种新的导频提取算法，对10个ADALM-PLUTO软件定义无线电设备(SDR)辐射出的信号提取其导频，并建立了3种不同距离条件下的导频数据集。提出将Inception网络结构用于射频指纹识别，在10 m无线传输距离下达到了98.58%的分类精度，相较于现有基于AlexNet网络改进的卷积神经网络(CNN)，分类精度有所提升。 

A new RF fingerprint identification method based on preamble of signal

Deep learning-based RF fingerprint recognition methods now primarily use raw data samples as the input of the network, never taking into account how the signal's content affects classification outcomes, and the structure of the network is relatively simple. In response to the above problems, the preamble of the signal as the input of the network was studied and we proposed a new preamble extraction algorithm.We extracted the preamble of 10 ADALM-PLUTO software-defined radios (SDR) and built the preamble data sets at three different distances.The Inception network structure is proposed to be used in RF fingerprint identification in this paper, and the classification accuracy is still 98.58% under the wireless transmission distance of 10 m. The classification accuracy is increased as compared to the pre-existing convolutional neural network (CNN) built on the AlexNet network.