Fingerprint Feature Extraction And Identification Of Radar Emitter

Specific emitter identification is the key to obtain important strategic information of the enemy and grasp the trend of war,so it is of great significance in EW countermeasures.The basis of specific emitter identification is the fingerprint features of the signal,that is,the parasitic modulation caused by the internal components of the emitter due to process reasons,improper maintenance,aging and other reasons.With the development of electronic technology,various new radars have emerged one after another,and the external electromagnetic environment has become more complex.How to accurately extract the fingerprint features of the signal and correctly identify the individual emitter is very important.Therefore,this paper first carries on the unintentional modulation modeling of the radiation source,and focuses on two aspects: feature extraction and individual recognition.Finally,a specific emitter identification method based on spectrum asymmetry is proposed,which is verified by simulation with the measured data of five emitters.The main research contents of this paper are as follows:Firstly,the basic structure of radar transmitter is introduced,the generation mechanism of unintentional modulation is analyzed,the unintentional modulation generated by oscillator and amplification chain in the transmitter is studied.It is found that the main form of unintentional modulation is phase noise.Then,the mathematical models of unintentional modulation of LFM signal and BPSK signal are established respectively,and the simulation experiments are carried out with three radiation sources with different phase noise.The frequency spectrum and power spectrum reflect the differences of output signals of different radiation sources.Secondly,the methods of fingerprint feature extraction and individual recognition are studied.The feature extraction method based on signal envelope,wavelet transform and1.5-dimensional spectrum is studied,and the Relief-F algorithm is used to optimize the energy spectrum features after wavelet transform,eliminate the invalid features,and improve the effectiveness of the features;Aiming at the problem that the feature dimension of1.5-dimensional spectrum is too high,the KPCA method is used to reduce the feature dimension,remove the redundant information in the feature,and only retain the representative feature information in the feature.Then the hybrid optimization of ABC-PSO is studied to optimize the parameters of SVM,and the optimized SVM is used to classify the envelope characteristics,wavelet transform energy spectrum characteristics and 1.5-dimensional spectrum characteristics of signal.Through the simulation verification of the collected three emitter data,under the LFM signal types,when the signal-to-noise ratio is 5d B,the correct recognition rate of the three methods can reach more than 95%,and then verify the generalization of the algorithm.Finally,one-dimensional convolutional neural network classifier is studied to classify and recognize 1.5-dimensional spectral features.When the signal-to-noise ratio is 5d B,the recognition rate can reach 93.5%.Finally,according to the phenomenon that the addition of phase noise will cause the spectrum asymmetry of LFM signal,a radar emitter individual recognition method based on spectrum asymmetry is proposed,and the emitter individual recognition is realized by evaluating the spectrum asymmetry.Through the simulation experiment of five groups of individual radiation sources,it is found that the performance of this method is stable,and the correct recognition rate is 95.83% even when the signal-to-noise ratio is-5dB.