Research On Multi-antenna Emitter Identification Using Power Amplifier Nonlinearity

In recent years,the rapid development of communication technology has spawned a prosperous scene in the mobile terminal market,and there are various mobile devices are emerging in endlessly.The proposal of 5G network is a prelude to the era of "Internet of Everything".The number of wireless mobile terminals will show explosive growth.In particular,terminal devices supporting Multiple Input Multiple Output(MIMO)will dominate the future network Status,so the identification capability for multi-antenna communication devices is essential in the field of wireless network security.Specific Emitter Identification(SEI)technology can realize the classification and recognition of different wireless emitters by extracting the radio frequency fingerprint of the physical layer,so as to detect illegal intrusion units and ensure the security of the system.However,at present,there is little open research on multi-antenna SEI,and in the actual communication environment,wireless channels have a greater impact on the stability of RF fingerprint features.In order to solve these problems,this paper takes the parameter features of nonlinear model of power amplifier(PA)inside the multi-antenna communication radiation sources as the research object,the coefficients of PA model and channel gain are jointly estimated in the MIMO communication system to develop the research of multi-antenna SEI under the influence of wireless channels.In this paper,based on the SEI processing flow of feature extraction and classification,the research work can be summarized into the following aspects:1.In the MIMO communication system with training sequence assistance,as for the impact of multipath channel in the actual environment,this paper uses a memoryless polynomial model to fit the nonlinear behavior of the PA inside the multi-antenna radiation source,and constructs a system of equations for the parameters to be estimated with respect to the multi-antenna observation signal with the help of all training sequence.Then,in this paper,two approaches to jointly estimate the coefficients of PA model and multipath channel gain are proposed from the perspective of solving linear least squares(LLS)and non-linear least squares(NLS)problems,respectively.Besides,this paper also gives a theoretical analysis of the applicable conditions of the proposed method.Further,in order to avoid the problem of pre-estimating channel order and improve the practicality of the approaches,this paper extends the feature extraction algorithm based on training sequence assistance to Orthogonal Frequency Division Multiplexing(OFDM)communication system,and the effectiveness of the algorithm is verified through physical experiments.2.As for the problem of insufficient training sequences in the actual system,in order to reduce the overhead of training sequences in the feature extraction process,this paper takes unknown data sequence as hidden variables and proposed a semi-blind joint estimation method of coefficients of PA model and channel gain based on the framework of Expectation Maximization(EM)algorithm.At the same time,this paper also introduces a priori hypothesis that the channel follows Gaussian distribution to constrain the iteration process of the proposed algorithm,and the simulation results show that it improves the performance of the algorithm.3.In the stage of classification and recognition,for the estimated parameter features of PA model,this paper uses a classifier based on the minimum error probability criterion to identify multi-antenna radiation sources,and analyzes the average error rate of the classifier through theoretical derivation,which provides a theoretical guide for the simulation experiment results.