Research On Key Technologies Of Communication Emitter Identification

With the rapid development of communication technology,the types and quantities of communication equipment are increasing rapidly,which brings more application requirements and greater technical challenges to the identification and authentication of communication emitters.In recent years,many advances have been made in the research of specific emitter identification(SEI),which has gradually become a research hotspot in the field of communication.In the unified processing flow,this thesis focuses on high-precision preprocessing methods,feature extraction methods based on expert experience,feature extraction methods based on deep learning,individual identification methods considering Doppler effect,and individual identification methods for time-division multiple access(TDMA)signals.The main contributions and innovations are as follows:1.For the preprocessing of demodulated data,a high-precision symbol synchronization algorithm based on two-stage refinement is proposed.Firstly,the main problems of the existing methods in SEI scenarios are pointed out.Then,the existing problems are soloved by using the interpolation filter based on window method and the time delay estimator based on forward decision feedback.The experimental results show that,this method effectively improves the processing accuracy of interpolation correction and time delay estimation,and the recognition gain of the signal is more than 2% compared with the existing methods.2.For the preprocessing of non-demodulated data,a high-precision preprocessing scheme based on interference cancellation is proposed.Firstly,the signal model of interference cancellation is established.The distorted signal is taken as the research object,and the ideal signal is regarded as interference.Then,a high-precision preprocessing scheme based on interference cancellation is given.Finally,the influence of processing accuracy on the cancellation effect is analyzed by using the theory of interference cancellation ratio.The experimental results show that,the proposed method can eliminate the adverse effects of transmitted symbols on the stability of features with high accuracy,and effectively improve the performance of SEI compared with the existing methods.3.Using expert experience,a feature extraction algorithm based on peak-to-average power ratio(PAPR)is proposed.Firstly,the relationship between PAPR and power amplifier nonlinearity is analyzed,and then the statistical features of numerical value and histogram are extracted from PAPR vector.This method is simple and feasible,and is very suitable for engineering practice.The experimental results show that,this method is more specific to the non-linear characteristics of power amplifier,and the recognition gain is about 10% under simulation and actual conditions compared with the existing methods.4.Based on expert experience,a feature extraction algorithm based on signal trajectory decomposition is proposed.Firstly,the distortion models of modulator,filter,oscillator and power amplifier are given based on the transmitter structure,and the visual representation of various distortions on signal trajectory image is deduced and analyzed.Then,the signal trajectory is defined and mathematically described,and the feature extraction is realized by trajectory decomposition.The experimental results show that,this method can extract a variety of distorted fingerprint features jointly,and significantly improve the performance of SEI.The recognition gain under simulation and actual conditions is about 30% compared with the existing methods.5.Using deep learning technology,an individual recognition algorithm based on Hilbert spectrum is proposed.Firstly,the received signal is transformed by Hilbert-Huang transform,and the grayscale image of Hilbert image is used as signal representation.Then,the deep residual network is constructed to learn and extract individual visual differences in the image directly.The experimental results show that,this method can overcome the limitations of existing expertise,effectively improve the recognition performance compared with the existing methods,and has good adaptability in complex communication systems and complex channel environments.6.Using deep learning technology,we propose an individual recognition algorithm based on signal trajectory image.Firstly,the random interference of irrelevant factors is eliminated by preprocessing;secondly,on the basis of theoretical analysis,the grayscale image of signal trajectory image is selected as signal representation,which can effectively present a variety of distorted visual features;finally,the joint extraction of a variety of distorted features is realized by using the deep residual network.The experimental results show that,this method has both high information integrity and low computational complexity,and significantly improves the performance of SEI compared with the existing methods,and has good practicability for actual signals.7.Considering Doppler effect,a new feature extraction method based on improved signal trajectory decomposition is proposed.Firstly,the signal model under Doppler channel is established;secondly,the influence of Doppler effect on the visual features of the signal trajectory image is analyzed;finally,a variety of stable fingerprint features are defined and extracted.The experimental results show that the algorithm is effective and stable compared with the existing methods,and effectively improves the SEI performance under the Doppler channel.8.A new radio frequency feature and processing flow is proposed for the TDMA system.Firstly,based on transmitter structure,the generation mechanism of features is analyzed with carrier phase and the extraction method is given.Secondly,the detection statistics are constructed with the features,and the adaptive threshold is deduced to realize the user identity detection in adjacent slots.Finally,a new processing flow is designed by data accumulation according to the detection results,which breaks the traditional thinking of identifying each slot separately.The experimental results show that the new process can effectively improve the performance of SEI compared with the traditional methods.