Active Jamming Recognition Based On Deep Learning Under Small Sample Condition

Radar is an integral part of modern warfare.With the rapid development of electronic countermeasure technology,the environment faced by radar is increasingly complex.In addition to the required echo signal,the received signal of the radar often contains a lot of interference.Due to the existence of these interferences,it is extremely difficult for radars to track and detect targets.Therefore,in order for the radar to be able to effectively identify the target,it is necessary to effectively classify and identify the received interference.However,due to the particularity of radar,we cannot obtain a large amount of data for relevant feature extraction and identification and classification.In this context,this topic is aimed at simulation and mathematical modeling of different types of radar interference,and in the follow-up Data augmentation under sample conditions,support vector machines and convolutional neural network methods are used to classify different radar interference under small sample conditions.The main research contents include:(1)Common interferences include blocking interference,ai ming interference,function sweep interference,intermittent sampling and forwarding interference,smart noise interference,dense false target interference,distance towing interference,and composite interference smart noise and dense false target interference.Simulation,and verify the interference after simulation,provide a basis for subsequent interference classification and identification.(2)Learn the relevant knowledge of the existing classifier models,including support vector machines and convolutional neural networks,and establish related classifier models to facilitate the classification of radar interference under the condition of subsequent small samples.(3)Aiming at the interference data in the case of small samples,this paper uses a generative countermeasure network to augment the data,and on the basis of the original generative countermeasure network,three improved generative countermeasure networks are used,including: WGAN-GP,CGAN,CGAN.In the follow-up experiments,we observe the generation effects of the three networks by comparing the original data and the time domain map and the amplitude normalization map of the data generated by the three generation confrontation networks.(4)The classification of radar interference data is carried out by using the support vector machine classifier and the convolutional neural network classifier under the condition of small samples,and the pros and cons of the two classifier models are obtained according to the final classification accurac y.In the subsequent experiments,the data generated by the three generation confrontation networks and the original data are added as a new training set,and the subsequent recognition accuracy rate is observed under the condition of the test set unchange d,and the final recognition rate is obtained by comparing The use of generative adversarial networks for data augmentation under small sample conditions can improve the recognition accuracy of the final interference.