Research On Radio Frequency Fingerprint Identification Method For Linear Frequency Modulation Signal

The linear frequency modulation(LFM)signals have been widely used in communication,radar,sonar and other fields based on their characteristics that they have a simple signal form and can improve the detection ability and anti-interference ability of radar systems through pulse compression technology.As the LFM signal is used by more and more wireless devices,it is of great significance to recognize the emitter of the LFM signal.Traditionally,the identification of the emitter of the LFM signal is mainly based on the pulse parameter characteristics,including the pulse period and the pulse amplitude,etc.However,this identification method is vulnerable to spoofing or forgery attacks,which ultimately leads to insufficient identification accuracy.In contrast,the method based on radio frequency fingerprint identification technology is to identify the emitter by extracting the characteristics of the physical layer of the emitter carried by the LFM signal.These physical layer characteristics are produced by the non-ideality of the emitter,so they are unique,persistent and not easy to be cloned.By extracting radio frequency fingerprints,the emitter can be uniquely identified.Therefore,this thesis mainly studies the LFM signal emitter identification based on radio frequency fingerprint identification technology.This thesis analyzes and researches on the methods and steps of radio frequency fingerprint identification technology for the LFM signal,and designs a scheme for identifying radar emitter based on radio frequency fingerprint identification technology.The main work and innovations of this thesis are as follows:In order to improve the signal-to-noise ratio(SNR)of LFM signals and reduce the influence of noise interference on the identification of LFM signal emitters,this thesis analyzes and compares the advantages and disadvantages of traditional noise reduction algorithms based on empirical mode decomposition,and proposes a LFM signal noise reduction algorithm based on window function and empirical mode decomposition.The algorithm takes full advantage of the approximate stability of the LFM signal in a short period of time,and combines the adaptive and multi-resolution advantages of the empirical mode decomposition algorithm.Therefore,in the case of low SNR,the algorithm is simple to implement and has a good noise reduction effect on LFM signals.According to simulation,when the SNR is in the range of 0d B to 20 d B,compared with the three traditional noise reduction algorithms based on empirical mode decomposition,the noise reduction algorithm proposed in this thesis obtains a higher SNR of the noise reduction signal? When the SNR is 25 d B and 30 d B,although the SNR of the noise reduction signal obtained by the noise reduction algorithm proposed in this thesis is slightly lower than the SNR of the noise reduction signal obtained by the empirical mode decomposition combined with the wavelet packet decomposition noise reduction algorithm,the noise reduction algorithm proposed in this thesis requires fewer parameters to be adjusted during the optimization process.In order to overcome the defect that the classification accuracy of single transient feature is not high enough in LFM signal emitter identification,this thesis explores the effect of fusion features on the LFM signal emitter identification accuracy.By comparing the classification results of the fused transient feature and the single transient feature,it can be obtained that the fused transient feature has a higher classification accuracy.According to simulation,among the single transient features extracted,the wavelet packet node energy features have the highest average classification accuracy.When the SNR is 25 d B,the average classification accuracy of the wavelet packet node energy feature under the random forest classifier reaches 92.37%.After the feature fusion of the four features of instantaneous amplitude,instantaneous phase,Fourier spectrum and wavelet packet node energy,the average classification accuracy of the obtained fused feature reaches 96.74%.Based on the above research on the denoising algorithm of combined window function and empirical mode decomposition as well as the fusion feature extraction algorithm,this thesis designs a scheme based on radio frequency fingerprint identification technology to identify radar emitter.By simulating the received 8 groups of radar radiator signals,the method and steps of radio frequency fingerprint identification for radar radiator signals are explained in detail.Finally,when the SNR is 30 d B,the classification accuracy of the test set of the radar radiator signals reaches 97%.