Research On Recognition Methods For Typical Deception Jamming Of Terminal Guidance Radar

In the final flying stage of a missile,the terminal guidance radar often faces various types of jamming.In this case,different suppression method is needed in the process of antiinterference.Therefore,identifying the type of jamming faced by the terminal guidance radar correctly is the prerequisite for radar countermeasures.In some way,the effectiveness of interference identification directly determines the success of anti-jamming,which has vital impact on modern warfare.The deception jamming based on digital radio frequency storage mode is wildly used in modern warfare.This type of jamming is quite flexibility and the parameters are easy to change.More importantly,the deception jamming signal is highly similar to the target echo.Because of the similarity of jamming signal and target echo,it is often difficult to recognize the type of signal.To solve this problem,the thesis studies the recognition methods of several typical active deceptive jamming and simulates the jamming signals.After that,the thesis extracts the feature of different type of jamming and combines machine learning method with neural network to identify the type of interference.Firstly,the thesis studies the method using image process according to the difference of timefrequency spectrum between various types of interference.The method uses gray scale image of time-frequency spectrum to calculate the gray level co-occurrence matrix and extract texture features to identify the interference with support vector machine.Secondly,the empirical mode decomposition method is used to decompose the mixed signal of target and jamming to obtain the intrinsic mode function of the signal.After selecting the decomposed components,the waveform entropy of the intrinsic mode function is extracted to classify the interference.Then the two types of features using different extraction methods are combined to construct a high-dimensional joint feature and SVM is used to recognize the signals.The simulation results show that the recognition effect of joint feature is better than that of a single type feature.Finally,aiming at the problem of insufficient sample in real world,the thesis uses Siamese network and prototype network combing metric learning in jamming identification.The simulation results show that even with a smaller number of training samples,the recognition rate using neural network is greatly improved than using SVM.To solve the problem of recognizing new type of jamming,the thesis uses Siamese network to identify the unknown type of signal.The simulation results show that Siamese network performs well in recognizing unknown type of signal.The interference recognition method studied in the thesis is quite valuable in theory and practical engineering.