基于ML-Decoder多分量雷达信号脉内调制识别方法

A Multi-Component Radar Signal Intra-Pulse Modulation Recognition Method Based on ML-Decoder

在现代电子侦察领域,由于电磁环境复杂,脉冲流密度较大,存在同时接收多个雷达信号的情况,多个雷达信号会在时域和频域出现重叠问题,使得雷达信号的特征变得混淆复杂。雷达信号的脉冲调制识别研究在单分量信号中取得了较好的效果,而在多分量雷达信号领域中,需要更多创新方法。为了解决上述问题,提出基于多标签解码器网络(Multi-Lable Decoder Network)框架。该网络框架首先用Choi-Williams分布(Choi-Williams Distribution,CWD)将一维信号转变为时频图。然后通过卷积神经网络提取特征,将提取的特征和查询向量一起送进decoder分类器中。decoder分类器通过标签查询的方法匹配特征信息,有效地避免传统卷积神经网络通过全局池化而淹没丰富的特征。用该方法对由六种典型雷达信号随机组成的多分量雷达信号经行调制识别分析,平均识别准确率达到93.9%,优于所对比的其他深度学习算法。

In the field of modern electronic reconnaissance,due to the complex electromagnetic environment and high pulse current density,multiple radar signals may be received at the same time,and multiple radar signals will overlap in the time domain and frequency domain,making the characteristics of the radar signals confusing and complex.The research on pulse modulation recognition of radar signals has achieved good results in single-component signals,but in the field of multi-component radar signals,more innovative methods are needed.In order to solve the above problems,a multi-label decoder network framework is proposed.Firstly,the Choi-Williams distribution(CWD)is used by the network framework to transform the one-dimensional signal into a time-frequency graph.Then features are extracted through a convolutional neural network,and the extracted features and query vectors are sended to the decoder classifier.The decoder classifier matches feature information through label query,effectively preventing traditional convolutional neural networks from drowning rich features through global pooling.This method is used to perform modulation identification analysis on multi-component radar signals randomly composed of six typical radar signals.The average recognition accuracy reaches 93.9%,which is better than other deep learning algorithms compared.