Research On Radar Active Jamming Recognition Method Based On Deep Neural Network

With the improvement of science and technology,radar systems are playing an increasingly important role in modern electronic warfare.Radar active jamming can effectively interfere with the normal operation of the radar system.Therefore,the radar system inevitably needs to have the ability to automatically identify the jamming signal,so that the appropriate anti-jamming means can be selected according to the type of the jamming signal.In recent years,deep learning has received widespread attention,among which the AlexNet model has made many achievements in image processing,pattern recognition and other related fields.Based on the success of the AlexNet model in deep learning in image recognition,this paper proposes a radar jamming signal recognition method that combines time-frequency analysis and AlexNet migration deep learning,and improves it under low interference-to-noise ratio.Firstly,the common radar active jamming signal is modeled and simulated,and its principle is analyzed.Then the interference recognition algorithm based on decision tree and neural network is briefly introduced,and their advantages and disadvantages are analyzed.Secondly,several time-frequency analysis methods are simulated.By analyzing and comparing the simulation results,the CWD transform is selected to perform time-frequency analysis on the radar active jamming signal,and 7 kinds of radar active jamming CWD time-frequency images are generated.Thirdly,the interference recognition algorithm based on AlexNet migration deep learning is fully implemented in the MATLAB platform,which includes the generation of data sets and the determination of network training parameters.On the basis of algorithm identification,the interference is analyzed under different interference-to-noise ratios.Experimental results show that the recognition rate of radar jamming signals can reach 98.7% under the full interference-to-noise ratio(-10 d B?20d B),and the recognition rate is low under low interference-to-noise ratio.Finally,an improved AlexNet model is proposed,which uses Swish-Re LU6 activation function,Adam optimizer,and adjusts its network structure,and compares the recognition rate with the original AlexNet model at low interference-to-noise ratio.The simulation results show that the improved AlexNet model has a higher recognition rate under low interference-to-noise ratio(-10 d B?0d B),and the overall recognition rate of interference with interference-to-noise ratio at-10 d B reaches95.7%,which is an increase of 5.5%.