Research On Modulation Recognition Algorithm Of Low Probability Of Intercept Radar Signal

Low Probability of Intercept(LPI)radar uses special transmitting signals to prevent noncooperative receiver from intercepting and detecting its transmitting signals.This kind of signal has the advantages of high concealment and strong anti-jamming,and it is widely used.Identification of modulation mode of LPI radar signal is one of the important means to obtain non-cooperative radar information.Only by grasping these information,can we make appropriate decisions and conduct targeted jamming,deceiving and attacking non-cooperative radar.Therefore,it is of great significance to study the modulation mode identification of LPI radar signal.In this paper,a total of 12 LPI radar signals including Linear frequency modulation(LFM),Costas signal,Binary Phase Shift Keying signal(BPSK),polyphase code signals and polytime code signals are studied.The algorithm of image feature extraction and deep learning was used to identify the modulation mode,and the recognition rate was high under the condition of low SNR.The main contents are as follows:1.For the recognition of LFM,Costas,BPSK,Frank and polytime code(T1,T2,T3,T4)signals,a method of artificial feature extraction based on time-frequency images is proposed.This method extracts the Choi-Williams Distribution(CWD)image of the signal,from which it can be found that the time-frequency images of different LPI radar signals have obvious differences.Then the images are preprocessed by means of grayscale,adaptive Wiener filter,binarization and bicubic interpolation.Then Hu moment and pseudo-Zernike moment are used to extract the shape features of the image.Local Binary Pattern(LBP)and Enhanced Local Derivative Pattern(ELDP)are used to extract the texture features of the image.Finally,the treebased machine learning process Optimization Tool(TPOT)is used to select the most suitable classifier and the corresponding parameters to classify the LFM,BPSK,Costas,Frank code signals and polytime code signals.The simulation results show that the method used in this paper has a good effect on the identification of 8 LPI radar signals under the condition of low SNR.2.The Single Shot Multibox Detector(SSD)model was introduced to solve the timeconsuming and laborious problem of artificial feature selection of LPI radar signals in the case of low SNR and small samples.Firstly,the edge image and four-direction texture feature image of CWD time-frequency image are extracted by using Sobel operator and ELDP algorithm,and the eight-channel image is obtained by using them to enhance the original image.Then CSPRes Ne Xt model and Focal Loss were used to replace the backbone network and the confidence Loss in the Loss function of SSD respectively.Finally,the Convolutional Block Attention Module(CBAM)module is introduced into the model and the process of image enhancement is reduced,thus the Improved Single Shot Multi Box Detector(ISSD)model is obtained.3.For the recognition of LFM,BPSK,COSTAS,polyphase code(Frank,P1,P2,P3,P4codes)signals and polytime code signals,a method based on ISSD model is proposed.Firstly,the CWD time-frequency images of the signals are extracted,and then the images are preprocessed by adaptive Wiener filter and bisubic interpolation.Then,12 LPI radar signals are detected and identified by ISSD model.Experimental results show that this method can increase the recognition rate of LPI radar signal in the case of low SNR and small sample.