Research On Radar Jamming Signal Recognition Method Based On Deep Learning

With the development of electronic information technology,modern war has gradually evolved into a high-tech war dominated by electronic warfare.Among them,radar,as an important component of electronic warfare,is the key to obtain complex battlefield information.In order to obstruct enemy radar access to important information,radar jamming technology and jamming equipment have been developed.Therefore,in order to ensure that the radar can fully play its role in the battlefield,higher requirements are put forward for the anti-jamming performance of radar.As the first step of radar anti-jamming process,accurate recognition of radar jamming signal is very important for radar survival.Aiming at the shortcomings of traditional radar jamming signal recognition methods,this paper explores the effectiveness of deep learning-based methods in the field of radar jamming signal recognition.As an important component of deep learning,the methods based on convolutional neural network(CNN)show a certain ability in the field of extracting discriminative features and accurate recognition.In order to give full play to the potential of deep learning,this paper proposes radar jamming signal recognition methods based on CNN,and conducts experimental verification on 12 types of typical radar jamming signals.The main research contents in this paper are listed as follows:Firstly,the theory and model design process of deep CNN are studied thoroughly,and the radar jamming signal recognition model based on 1D-CNN is proposed according to the characteristics of radar jamming signals.Furthermore,in order to improve the generalization ability of the recognition model,this paper proposes a soft label smoothing method and applies it to the design process of the recognition model.Compared with other typical radar jamming signal recognition methods,the proposed method achieves the optimal recognition performance on the radar jamming data set in this paper.Secondly,the acquisition and annotation of radar jamming signal is a difficult and tedious process in the actual battlefield.Therefore,the number of samples for training deep learning models is often limited,and the lack of training samples easily leads to the phenomenon of overfitting of deep models.In order to solve the problem of insufficient jamming samples,the radar jamming signal recognition model based on deep siamese network is proposed in this paper.Compared with other typical radar jamming signal recognition methods,the proposed method achieves the optimal recognition performance when the radar jamming training samples are limited.Finally,in order to further improve the robustness and recognition performance of radar jamming signal recognition model,the fusion recognition models of radar jamming signal at feature and decision level based on CNN are proposed respectively.Compared with the originally designed radar jamming recognition model based on 1D-CNN and other typical radar jamming signal recognition methods,the CNNbased radar jamming data fusion recognition model has been improved significantly in terms of robustness and radar jamming recognition performance.