Suppression Of Compound Jamming Based On Jamming Recognition

With the continuous development of modern electronic warfare technology,the continuous emergence of various new types of composite jamming poses a great threat to the survival and operation of radars.Therefore,the performance of suppressing radar composite jamming is an important manifestation of radar anti-jamming ability,and it is also a hot spot in radar research.Aiming at the problem of main and side lobe coexistence compound jamming,the existing research work mainly adopts the spatial filtering method,but it cannot deal with the situation that the jamming and the target signal are incident from similar angles.The introduction of deep learning to extract jamming features provides a powerful technical means for this research field.Based on jamming cognition,combined with the space,time,and frequency information of jamming signals,this thesis proposes a joint suppression method based on orthogonal projection,transfer learning,and jamming reconstruction and cancellation.In this thesis,a two-step suppression process is adopted.First,the direction estimation combined with orthogonal projection is used for the sidelobe jamming,which not only suppresses the sidelobe jamming,but also fully retains the main lobe signal for the second step of processing.Then,the main lobe jamming characteristics are identified through transfer learning,and then the jamming signal is reconstructed and canceled to suppress the main lobe jamming;finally,the jamming is suppressed and the target is detected.The main contributions of this thesis are as follows:1.The typical suppression jamming and deception jamming models are analyzed,and based on this simulation,a single jamming data set is established to provide a basis for subsequent jamming identification.In addition,the uniform linear array is used as the receiving model to establish a signal model in which the main lobe receives a spoofing jamming signal and the side lobe receives at least one suppressing jamming signal;2.Aiming at the problem that the method based on the blocking matrix will weaken the desired signal close to the jamming,the direction estimation combined with the orthogonal projection is used to suppress the sidelobe,and the main lobe signal is reserved for subsequent processing.Simulation experiments show that this method can suppress multiple sidelobe jammings of different styles,and the jamming-to-noise ratio of the sidelobe is reduced to about 0dB after suppression.In addition,the orthogonal projection method solves the common problems of main beam deformation and sidelobe level increase in the blocking matrix method,and does not lose the degree of freedom of the received signal;3.Aiming at the problem that when the model is trained with a complex jamming dataset,the model will learn features that suppress the jamming,and a transfer learning method is used.The composite jamming data set is established with the signal after sidelobe jamming suppression.Using single jamming as the source domain and composite jamming as the target domain enables the model to focus on learning the jamming characteristics of the main lobe without being affected by suppressing jamming,and adapts well to new tasks,achieving a recognition rate of 90.95%.In addition,the interpretability analysis was carried out through the Integrated Gradients method,which proved that the model extracted the jamming features well;4.After identifying the pattern of mainlobe jammings,parameter estimation is performed by time-frequency domain analysis to reconstruct the resulting jamming signal.Combined with the spatial spectrum estimation method,the array steering vector is reconstructed,and then the received jamming signal is reconstructed to cancel the jamming signal and suppress the main lobe jamming.For the two kinds of spoofing jamming,the signal-to-jamming-noise ratio is increased by 6.51 dB and10.31 dB respectively.By introducing the method of reconstruction and cancellation,the problem of suppressing the echo of the target due to the close proximity of the target and the jamming in the traditional method is avoided,and the adaptability of the radar is enhanced.