Fingerprint Feature Extraction Of Radar Signal And Individual Identification Of Radiation Source

Individual identification of radar radiation sources is an important part of the electromagnetic spectrum warfare,and the control of the electromagnetic spectrum plays a very important role in the battlefield.How to extract stable and effective fingerprint characteristics of radar signals from radar signals is the focus of current radar individual identification research.The main research contents of this article are as follows:1.This article constructs individual models of radar radiation sources from the perspective of phase noise and the nonlinear model of power amplifier.Taking the linear frequency modulation signal and the two-phase coded signal as examples,multiple radar radiation individuals are simulated according to the phase noise carried by the individual radar radiation source and the nonlinear distortion of the power amplifier of the individual radar radiation source,for subsequent analysis in this article.2.In view of the higher-order spectrum that can better characterize the signal information,this article proposes a bispectrum-based radar signal fingerprint feature extraction method to extract signal kurtosis,bispectral waveform entropy,bispectral zero slice box dimension,bispectral diagonal slice Integral mean,bispectral singular value entropy,bispectral maximum spectral kurtosis,bispectral energy entropy.Seven eigenvalues are used as the radar signal fingerprint eigenvectors,and then the XGBoost classification algorithm is used to complete the recognition of the radar radiation individual.Simulation experiments show that the method proposed in this paper can improve the individual recognition effect of radar radiation sources.3.In view of the too high dimension of the signal time-frequency feature map,it is not conducive to directly extracting the signal fingerprint feature.This article uses the two time-frequency domain transforms of radar signals to perform the secondary extraction of the fingerprint characteristics of the signal.The fingerprint feature vector is then formed to identify the individual radar radiation source.Marginal spectrum entropy is extracted from the radar spectrum's Hilbert-Huang transform as the radar signal fingerprint feature.The pseudo-Wigner-Ville transform is performed on the radar signal to extract the singular value entropy as the signal fingerprint feature.Then construct an effective two-dimensional signal fingerprint feature vector.And verify its effectiveness through identification experiments.4.This article proposes to use convolutional neural networks to identify individual radar radiation sources.The advantage of using a convolutional neural network is that it can automatically extract the fingerprint characteristics of the signal to complete the individual identification of the radar radiation source,avoiding the tedious of manually extracting the fingerprint characteristics of the signal.The simulation experiments show that when the signal-to-noise ratio is 5dB,the recognition rate of different radiators constructed by the power amplifier using the time-frequency map and convolutional neural network after pseudo Wigner-Ville transform reaches more than 90%.The recognition effect is better than the existing individual recognition methods of radar radiation sources.