基于卷积神经网络的雷达辐射源稳健识别方法

针对低信噪比下，基于传统统计特征的雷达信号识别方法对复杂调制信号类型识别性能不高，因而处理复杂度高的问题，提出一种基于卷积神经网络的雷达辐射源信号稳健识别方法。该方法通过提取信号的瞬时相位特征，获得变换域的表征信号，将其作为卷积神经网络的输入，实现雷达辐射源信号的快速识别。针对瞬时相位特征对于信噪比敏感的特点，采用主成分分析方法对信号特征域进行降噪处理，提升模型对噪声的稳健性。通过仿真实验验证了所提出方法在不同信噪比下对7种调制信号类型的识别性能，通过理论分析及不同方法的实验对比，验证了算法具有耗时较短、识别准确率较高、噪声稳健性好等优势，具有良好的工程实用性。

A robust radar radiation source recognition algorithm based on convolutional neural network

Aiming at the problem that the radar signal recognition method based on traditional statistical features has low recognition performance and high processing complexity for complex modulation signal types under low signal-to-noise ratio, this paper proposes a robust radar emitter signal recognition method based on deep neural network. This method obtains the representation of the signal in the transform domain by extracting the instantaneous phase characteristics of the signal, and uses it as the input of the deep neural network to realize the rapid recognition of the radar emitter signal. In view of the fact that the instantaneous phase feature is sensitive to the signal-to-noise ratio, the principal component analysis method is used to denoise the signal feature domain to improve the robustness of the noise model. Through simulation experiments, the recognition performance of the proposed method for seven modulation signal types under different signal-to-noise ratios is verified. Through theoretical analysis and experimental comparison of different methods, and it verifies that the proposed algorithm has the advantages of short time-consuming, high recognition accuracy and better noise robustness, which can be applied in engineering practicability.