Research On Intelligent Perception Method Of Radar Jamming In Complex Electromagnetic Environment

With the rapid development of modern electronic technology and artificial intelligence technology,a large number of intelligent technologies are applied in the radar electronic warfare field,and radar electronic countermeasures are becoming more and more intelligent.In the radar intelligent anti-jamming system,the jamming types and parameters are firstly obtained by the jamming perception module,and then the anti-jamming decision module generates the corresponding anti-jamming measures according to the feedback information of the jamming perception module.Therefore,the radar jamming perception module is equivalent to the "eye" of the intelligent anti-jamming system,and the jamming perception is the premise of the intelligent anti-jamming system to achieve better anti-jamming effect,which is very important to the radar anti-jamming.Aiming at the problem that the existing radar jamming recognition method based on deep learning is prone to overfitting when the sample size of the training set is small,the jamming recognition method in the few-shot scenario is studied.Two solutions are proposed:the first one is based on transfer learning.Finally,the pre-training model is used to learn transferable prior knowledge in a dataset containing a large number of samples,and then the prior knowledge is transferred to the jamming recognition model.The second idea is based on the idea of amplifying the amount of jamming samples,using the generative adversarial network to fit the distribution of each type of jamming data and the ability to generate differential samples,and design a few-shot jamming recognition model with a simple structure and easy training.Finally,the few-shot recognition model is evaluated in different few-shot scenarios,and the effectiveness of the small sample jamming recognition model based on transfer learning and generative adversarial network in few-shot scenarios is verified.Aiming at the problem that the existing jamming recognition methods in the open-set jamming environment will incorrectly identify the unknown jamming as the known jamming existing in the training set,this paper firstly conducts research on the jamming recognition method in the open-set scenarios and the open-set jamming recognition models based on confidence score and Open Max are proposed,respectively.The open-set jamming recognition model based on the confidence score first generates the probability distribution of the input jamming signal on the known jamming types through the jamming classifier,and then judges whether the jamming recognition result is reliable by calculating the confidence score of the probability distribution,if reliable output jamming classifier recognition results,otherwise output jamming for the unknown jamming.In order to solve the problem that the model based on confidence score needs to manually simulate the openset scenarios and set the confidence score threshold,this topic proposes an open-set jamming recognition model based on Open Max.Firstly,the jamming signal is input into the jamming classifier to extract the activation vector.Then,based on the activation vector of jamming samples and weibull distribution of each type of known jamming boundary,the Open Max layer containing the probability that the input jamming samples are unknown jamming types is constructed.Finally,the type with the highest probability in the probability distribution of Open Max output is selected as the recognition result output.The two open-set jamming recognition models are tested in different open-set jamming test environments,and the effectiveness of the open-set jamming recognition model in the open-set jamming scenario is verified.