基于卷积神经网络和模糊函数主脊坐标变换的雷达辐射源信号识别

针对人工提取雷达辐射源信号特征耗时长、特征不明显等问题，提出一种基于深度学习卷积神经网络和模糊函数主脊坐标变换的雷达辐射源信号识别方法。该方法通过快速离散分数傅里叶变换提取信号的模糊函数主脊，并将模糊函数主脊极坐标域的二维时频图作为卷积神经网络的输入，实现对不同雷达信号的分选识别。仿真实验结果表明：新方法不仅在信噪比为0 dB以上保持100%的识别率，在-6 dB时识别准确率也稳定在90%以上；相对于传统的雷达信号识别方法和其他深度学习模型识别方法，在识别率和鲁棒性上均有较大提升，具有一定的工程应用价值。

Radar Emitter Signal Recognition Based on Convolutional Neural Network and Coordinate Transformation of Ambiguity Function Main Ridge

For the time-consuming and inconspicuous features of artificially extracting radar emitter signal features，a recognition method based on deep learning convolutional neural network and coordinate transformation of ambiguity function main ridge is proposed. The proposed method is used to extract the main ridge of ambiguity function of signal through fast discrete fractional Fourier transform，and then take the two-dimensional time-frequency image of polar coordinate domain of ambiguity function main ridge as the input of the convolutional neural network to realize the sorting and recognition of different radar signals. Simulation experimental results show that the proposed method not only maintains 100% recognition rate above 0 dB，but also stabilizes the recognition accuracy rate above 90% at -6 dB. Compared with traditional radar signal recognition methods and other deep learning-based recognition methods，the proposed method has greatly improved recognition rate and robustness，and has certain engineering application value.