基于卷积神经网络的低截获概率雷达信号检测算法

为了解决雷达截获接收机对低截获概率(LPI)雷达信号检测效果不理想的问题，针对截获信号中有效信号脉宽长度来定义信号和噪声，该文提出一种基于卷积神经网络(CNN)的LPI雷达信号检测方法，利用卷积核与匹配滤波器结构上的相似性，在低信噪比下能够提高信号的检测准确率。利用大量的基于4种典型LPI雷达信号(线性调频信号(LFM)、非线性调频信号(NLFM)、二相编码信号(BPSK)、COSTAS频率编码信号)和白噪声信号的模拟数据集进行CNN模型训练，同时增加少量实测信号(LFM, BPSK)作为验证集进行适配，更好地拟合实测信号的检测模型。最终利用实际信号进行测试，实验结果表明:该文算法在低信噪比的情况下具有较好的检测效果，对多种调制方式、不同信噪比下的LPI雷达信号具有泛化能力。

Low Probability of Intercept Radar Signal Detection Algorithm Based on Convolutional Neural Networks

In order to solve the problem of radar intercepting receiver’s unsatisfactory detection effect on Low Probability of Intercept (LPI) radar signal, a method of LPI radar signal detection based on Convolutional Neural Networks (CNN) is proposed, which defines signal and noise by effective signal pulse width in intercepted signal. The similarity of the convolution kernel and the matched filter in the structure can improve the detection accuracy of the signal under the low SNR.A large number of analog data sets based on four typical LPI radar signals (Linear Frequency Modulation signal (LFM), NonLinear Frequency Modulation signal (NLFM), Binary Phase Shift Keying signal (BPSK), COSTAS frequency coded signal) and white noise signals are used for CNN model training. At the same time, a small amount of measured signals (Linear Frequency Modulation signal (LFM), Binary Phase Shift Keying signal (BPSK)) are added as verification set for adaptation, so as to match better the detection model of measured signals. Finally, the experimental results show that the proposed algorithm has a good detection effect in the case of low SNR, and has the ability to generalize the LPI radar signals under various modulation modes and different SNR.