Special Emitter Identification Via Deep Learning

It is very important to apply the specific emitter identification(SEI)techniques to improve the security of wireless communication systems and the ability of military communication countermeasure.Recently,the research and practical application of deep learning technology has developed rapidly,and it has achieved great success in the field of speech and image recognition.Its layered feature expression ability and unsupervised feature extraction ability are very conducive to the feature extraction of specific emitter.The deep learning model is designed to extract multi-dimensional and multi-level individual characteristics of specific emitter and identify them.In this paper,the distortion mechanisms of specific emitter are analysed and modeled.Convolutional neural network(CNN)is adopted to extract subtle features from the higher order spectrum estimation and the timefrequency transform of the received steady-state signals.The main works in this thesis are as following:1.Analysis and modeling of communication emitter distortion mechanism.This paper analyzes the distortion mechanism of the main devices in the process of communication radiation source signal from modulation to emission and establishes a mathematical model.2.Modulation recognition of communication emitter based on deep neural networks.In view of the common problem of digital modulation and analog modulation signal modulation recognition,the convolution neural network and recurrent neural network are established respectively to identify each modulation by using the IQ and amplitude-phase time series data of communication signals.Simulation results show that the algorithm presented in this paper has excellent performance.3.Special emitter identification via convolutional neural network and compressed bispectrum.In particular,the bispectrum of the received signal is calculated as a unique feature.Then,we use a supervised dimensionality reduction method to significantly reduce the dimensions of the bispectrum.Finally,a convolutional neural network is adopted to identify specific emitters using the compressed bispectrum.Both the simulations and the experiments validate our conclusion that the proposed approach outperforms other existing schemes in the literature.4.Special emitter identification via convolutional neural network and compressed time-frequency spectrum of Hilbert-Huang Transform(HHT).the Hilbert-Huang transform of received signals is calculated as a unique feature.Then,max-pooling technology and a supervised dimensionality reduction method are used to significantly reduce the dimensions of the original HHT.Finally,a convolutional neural network is adopted to identify specific emitters using the compressed HHT.This approach essentially extracts overall feature information hidden in the original signals,which can then be used to improve identification performance.