Research On Radar Emitter Modeling And Structure Recognition Algorithm

With the development of various electronic technologies in electronic warfare,the radiation structure has become increasingly diversified.As one of the core components of the radar system,the identification of the radiation source structure is helpful to identify the performance and threat level of the non-cooperative radar system,so that further measures can be taken.At present,there are relatively few researches on the system modeling and structure recognition of radar radiation sources in China,so this research is of great significance.In this article,the frequency synthesizer and power amplifier in the radiation source are respectively constructed and verified,combined into four radiation source models with different structures,and the models are simulated to obtain signals of various modulation methods,the time-frequency image of the signal is preprocessed and used to train the convolutional neural network(CNN)model,make the model have the function of identifying the structure of the radiation source,and realize the inversion process from the signal to the structure of the radiation source.Finally,the stability of the recognition model was verified under different signal-to-noise ratio(SNR).The main work content is as follows:Firstly,analyze the working principle and phase truncation error of the direct digital frequency synthesizer(DDS),construct its phase truncation model,and verify the model with the output signal spectrum.According to the components and working methods of the fractional frequency phase locked loop frequency synthesizer(Fractional-N PLL),a model with fractional spurs is constructed.In the modeling part of the power amplifier,the two most typical power amplifiers are selected: Solid-State Power Amplifier(SSPA)and Traveling-Wave Tube Amplifier(TWTA),the memory polynomial behavior model is used to simulate SSPA,and on this basis,the unique phase noise of TWTA is added to form the TWTA model.Two power amplifier models were tested,and the non-linearity and memory of the models were verified according to the AM-AM and AM-PM characteristic curves.Secondly,based on the structure and principle of the main vibration amplifier transmitter,four types of radiation source models are constructed,and according to the modulation principle of continuous wave(CW),linear frequency modulation(LFM),and binary phase shift keying(BPSK),adjust the frequency synthesizer model so that the four radiation source models can obtain CW,LFM,and BPSK signals respectively.Through the analysis of the signal spectrum,it is found that the spurious information carried by the signals of the radiation source models of different structures is unique.Finally,the simulation signals of the radiation source models of the four structures are subjected to short-time Fourier transform(STFT)to obtain their time-frequency images,construct a CNN-based radiation source structure recognition model,and use the preprocessed time-frequency images to train the CNN model and optimize the parameters.Perform recognition rate and robustness experiments on the trained model,when the SNR is-5d B,the recognition accuracy can still reach 87.2%,the stability of the identification system is further verified,the results show that the method can effectively identify the emitter structure and realize the inversion process from signal to emitter structure.