Research On Radar Active Jamming Recognition And Perception Method

Radar is the key equipment to perceive the electromagnetic environment,control the global information and compete for the electromagnetic power in the battlefield.In order to fully control the war process,the radar technology is developing gradually while various radar jamming technologies are also emerging one after another,which leads to the increasingly complex electromagnetic environment faced by radar equipment.Radar active jamming refers to the signal generated by jamming equipment.The active signal can enter the radar receiver and realize the purpose of jamming radar for target detection or destroying radar hardware system.If the normal detection and tracking function of radar is affected,it will be difficult for military personnel to obtain enough military information and make military decisions on the battlefield.Therefore,it is of great scientific value and significance to improve radar's recognition and perception ability to active jamming and enhance its antijamming ability.Aiming at the problem that the classifier used in radar active jamming recognition is not robust,a robust jamming recognition method is proposed in this paper.Existing methods usually adopt ensemble learning to improve the recognition probability for signals with different jamming noise ratios.However,when recognizing unknown signals,it is faced with the dilemma of how to select the best subclassifier,so it is not really suitable for signal recognition with different jamming noise ratios.Jamming recognition method proposed in this paper uses multiple JNR of the signal to train the single classifier.Because this method need not use the ensemble learning to train multiple classifiers,the training scale significantly reduced.Because of the joint learning of the characteristics of the signals with multiple jamming noise ratios,this method has a good recognition effect on the signals with different jamming noise ratios.The simulation results show that the recognition performance of the proposed method is better than that of ensemble learning in scenarios with different jamming noise ratios,and the algorithm achieves robustness and reduces computing overhead.In order to solve the problem that the artificial feature recognition effect is not good and too many features will cause "dimension disaster" in interference recognition,this paper studies and analyzes a variety of feature selection algorithms,improves them and integrates a variety of feature evaluation criteria.When the feature optimization algorithm is improved,the optimal feature subset is obtained by combining the feature selection process with the classifier training process to make the classifier more effective.The multi-evaluation criteria fusion feature selection algorithm comprehensively considers the correlation between different features of jamming signals and removes redundant information to achieve feature dimension reduction.Finally,several feature selection methods are evaluated through simulation experiments,and it is proved that the proposed algorithm has better performance and higher recognition probability.Aiming at the problems of excessive training cost and limited recognition accuracy in interference recognition in complex electromagnetic environment,a cascading classification and recognition method in multi-interference environment is proposed in this paper.When classifying and recognizing various interference signals in complex electromagnetic environment,a large data set will be obtained if unified analysis is carried out,and the training speed,recognition speed and recognition accuracy of the classifier will be affected if the data set is directly used for training.The method proposed in this paper firstly classifies various jamming signals roughly through a first-level classifier,then extracts characteristic parameters from the corresponding feature subset of specific jamming signals to train the classifier,and finally obtains specific signal types through a second-level classifier.The simulation results show that the proposed method can significantly reduce the computing cost and improve the training speed of the classifier and the probability of interference recognition.