基于卷积神经网络的雷达工作模式识别

Recognition of the Radar Working Pattern Based on Convolutional Neural Network

该文针对不同雷达工作模式的信号特征,提出一种基于卷积神经网络(convolutional neural network,CNN)的雷达工作模式识别方法。不同工作模式下的雷达信号的脉冲宽度、脉冲重复周期、脉内调制样式和数据率等特征均有所不同,所以该文利用这4个特征参数构建1个图像矩阵,再提取方向梯度直方图(Histogram of Oriented Gradient,HOG)的特征,送入CNN进行雷达工作模式识别。仿真结果表明,该识别方法有较高的识别准确率。

In this paper,we propose a working pattern recognition based on convolutional neural network(CNN)for radar signal characteristics in different Radar working patterns.The characteristics of radar signal,such as pulse width,pulse repetition period,intra-pulse modulation style and data rate,are different in the different radar working patterns.Therefore,we use these four features to construct an image matrix,extract the features of the Histogram of Oriented Gradient(HOG)and send them to CNN for radar working pattern recognition.Simulation results show that HOG+CNN has a high recognition accuracy.