Research On Cross-Receiver Specific Emitter Identification Method

Specific Emitter Identification(SEI)identifies the emitting device by extracting the subtle distortion contained in the signal.These distortions can express the characteristics of the emitter's hardware.SEI has been applied in the fields of wireless security,self-organizing networks,and military target recognition.In practical applications,scenarios such as wide-area moving target recognition and multi-platform cooperative identification are common.In these scenarios,the received data of multiple receivers are used to identify the targets,which is called cross-receiver SEI.In recent years,research on SEI technology has maintained a high popularity.However,most of the existing researches focus on performance in single-receiver scenario,and rarely discuss the influence of cross-receiver factors on recognition performance.In response to this problem,this article focuses on the cross-receiver SEI method.On the one hand,the influence of receiver distortion on the characteristics of individual identification fingerprints is quantitatively analyzed,and the cross-receiver SEI feature estimation method is obtained by improving the estimation model.On the other hand,to solve the problem of receiver distortion,receiver correction method based on migration learning and frequency domain adaptive filtering is proposed to further improve the reliability of cross-receiver SEI.The main contributions are as follows:1.Completed the quantitative analysis of the influence of receiver distortion on the fingerprint characteristics.Considering that the receiver processing is the inverse process of the emitter,the receiver distortion is modeled.The three types of SEI model,modulation distortion model,power amplifier distortion model,and filter model are analyzed.Then the receiver distortion mechanism model is used to derive the mathematics of the results under the influence of receiver distortion.According to the evaluation,the results of the modulation distortion model are relatively independent of the receiver distortion,while results of the power amplifier model and filter model are deeply influenced.2.An estimation method of SEI characteristics against receiver distortion is proposed.Based on the results of the quantitative analysis of receiver distortion,this paper improves the SEI method based on the mechanism model,and proposes an estimation method of SEI characteristics against receiver distortion.This method performs joint estimation of receiver distortion and fingerprint characteristics to obtain receiver distortion parameters and original fingerprints,and then acquires fingerprint that are independent of the receiver distortion.Simulation show that the fingerprint estimated by this method is effective in cross-receiver SEI.Experiments on actual equipment show that the offset of fingerprint features extracted by the improved method under cross-receiver conditions is less than 10% of the original method,and the accuracy in open-set recognition is improved by 8-10%.3.A cross-receiver SEI method based on transfer learning is proposed.The influence of receiver distortion on SEI can be seen as the distribution difference of fingerprint features.Therefore,we can learn from the methods of domain adaptation methods in transfer learning,and propose the use of transfer component analysis and joint distribution adaptation methods to correct the feature distribution.The actual equipment experiment shows that when 5 software radio equipment is used as the targets,the correct rate of the closed-set recognition is increased from 20%to more than 83%.4.An SEI receiver correction method based on frequency domain adaptive filtering is proposed.In order to unify the subtle distortions between different receivers and fundamentally solve the effects of receiver distortions,this paper proposes to use frequency-domain adaptive filtering(FDAF)to unify the subtle features of different receivers.Experiments on actual equipment show that after correction,the offset of the fingerprint feature of the radiation source between different receivers is well suppressed.When 5 software radio devices are used as the identification target,the accuracy rate is increased from 20% to 85%.At the same time,the influence of the receiver's signal-to-noise ratio and clock error on the correction effect is discussed.The results show that this method can be well used in the actual cross-receiver scenarios.