Research On Identification Approaches Of Novel Active Jamming For Radar

With the development of radar jamming technology,more and more active jamming modes with specific jamming effects have been proposed successively.Specially,with the emergence of novel active jamming with both high fidelity and complexity,the performance of radar has been challenged unprecedentedly,which has drawn great attention of the related researchers at home and abroad.In order to deal with the threat to the radar system caused by the novel active jamming,several jamming identification technologies are studied based on linear frequency modulation(LFM)signals commonly used in radar.Considering the practical application scenarios,this paper focuses on the theoretical analysis,feature extraction,and jamming identification methods of echo signals of jamming and target received by radar.The main work of this paper is:Firstly,this paper reviews the generation and mechanism of several new types of active jamming.Then,the mathematical models,including spectral dispersion interference,slicing combination interference,and intermittent sampling repeated and forwarded interference,are established.Meanwhile,the corresponding simulation and analysis are carried out in time domain and frequency domains.More importantly,the relationship between jamming and LFM signals is studied in this paper.Combined the generating principle of jamming signals with simulation results,it is able to be concluded that the spectral dispersion jamming,the slicing combination jamming,and the intermittent sampling and retransmission jamming all have dual jamming effects of deception and covering under certain conditions.Secondly,according to the time-frequency analysis results of jamming signals,both signal processing and image processing technology are combined effectively to find the characteristic differences among the novel active jamming.According to the shape and texture differences between jamming signals,the shape feature based on Zernike moment and texture feature based on gray co-occurrence matrix are extracted,and jamming signals are classified and further identified by support vector machine(SVM)based on the above two kinds of features.The experimental results show that the two kinds of time-frequency features have good performance in jamming classification and identification.Finally,depending on decision tree,SVM,andBP neural network classifier,the identification performance of mixed echo signals of jamming and target under different JNR conditions is simulated with fixed JSR of 5dB.Besides,when JNR is 5dB,the decision tree is used to identify the jamming signals.Specially,when JNR is greater than 5dB,the recognition rate of jamming signals can reach more than 95%.For SVM,the training data with JNR of 0dB,5dB,and 10 dB are leveraged for simulating and analyzing.In addition,under the condition of training data with JNR of 5dB and 10 dB,the simulation results show that when JNR reaches 5dB,the recognition rate of jamming signals reaches 100%.On the other hand,when JNR is 0 dB,the recognition rate of jamming signals can reach more than95% only when JNR is around 0 dB.At the same time,the classification and identification performance of the SVM under different kernel functions is analyzed and discussed,and the experimental results show that the jamming effect based on linear kernel function is better than that based on radial basis kernel function.Moreover,the BP neural network is introduced to identify jamming signals by using the training data with JNR of 5dB,and the recognition rate of jamming signals can reach more than 95% when JNR is 2dB.In addition,the classification performance for all three classifiers is carefully analyzed and compared,and the conclusion is drawn that the SVM is more suitable to identify several kinds of jamming studied in this paper.