Research On Radar Signal Recognition Based On Multi-window Spectrogram Analysis

Modulation recognition of radar signal plays an important role in reconnaissance system.However,with the gradual development of low probability of intercept(LPI)radar,the received signal-to-noise ratio(SNR)is getting lower and lower,and the traditional modulation recognition algorithm gradually fails.Modulation recognition algorithms of radar signal based on time-frequency image can solve the above problems to a certain extent,but the conventional time-frequency algorithm is seriously affected by noise.In order to solve this problem,this thesis studies LPI radar signal modulation recognition algorithm from the angle of improving time-frequency image,and proposes a time-frequency analysis method based on multi window theory.Combined with deep learning technology,this thesis proposes a modulation recognition algorithm of LPI radar single signal and modulation recognition algorithm of LPI radar multiple signals,and deeply discusses the performance of these two modulation recognition algorithms of radar signal.The main research contents include:Firstly,the existing modulation recognition algorithm of LPI radar signal based on time-frequency image often chooses to conduct in-depth research on the subsequent classification network,and uses the existing time-frequency analysis methods to generate the time-frequency image of the signal.However,these methods have low recognition rate and fuzzy image under low signal-to-noise ratio.In view of this problem,a multi window spectrogram algorithm is proposed to improve the quality of time-frequency image Compared with the conventional WVD(Wigner Ville distribution)and CWD(Choi Williams distribution)methods,the proposed method has better anti noise performance and time-frequency analysis performance,and the time-frequency image produced under low SNR has good quality.Secondly,from the perspective of deep learning technology,combined with Multi-window Spectrum and Image Net-VGG-f migration neural network,a modulation recognition algorithm of single LPI radar signal based on multi window spectrum is proposed.Above all,the Multi-window Spectrum algorithm is used to generate the time-frequency image of the signal.Then,the image is compressed,binarized and processed to meet the input requirements of neural network.Finally,Image Net-VGG-f migration neural network is used as classification network to realize the recognition of LPI radar signal modulation type.Simulation results show that the algorithm can effectively identify 12 kinds of LPI radar signals under the condition of low SNR.Finally,aiming at the problem that the conventional radar single signal recognition system fails to reach multiple signals at the same time,combined with the idea of blind source separation,this thesis proposes a multiple signals modulation recognition algorithm for LPI Radar Based on jade-mwsp.Firstly,the algorithm uses the blind source separation algorithm based on joint diagonalization of eigenmatrix(JADE)to separate the mixed signals,and finally transforms the multi signal recognition problem into single signal recognition The separated signals are sent to the single signal recognition system to realize the multi signal modulation recognition of LPI radar.In this thesis,a part of the dual signal combination is analyzed in Gaussian white noise.The simulation results show that the algorithm has good separation performance and recognition ability at low SNR.