Research On Individual Identification Technology Of Radar Emitter Based On Deep Learning

With the rapid development of equipment informatization and the increasingly complex electromagnetic environment of the battlefield,the emergence of new systems,new technologies,and multi-function radars have put forward a severe test for electronic warfare.The improvement of the consistency performance of the radar system makes it very difficult for the electronic reconnaissance system to accurately identify individual radar emitter.In order to better solve this problem,this paper will study from two aspects.On the one hand,it analyzes the reasons for the formation of fingerprint features of radar emitter,selects and extracts effective fingerprint features,and fuses them to form a multi-dimensional feature matrix.It is proposed to use the logistic regression algorithm and the gradient boosting decision tree algorithm to classify the feature matrix respectively,realize individual identification of radar emitter based on fingerprint characteristics.On the other hand,this paper proposes to combine stacked autoencoder and one-dimensional convolutional neural network with gradient boosting decision tree respectively,and the intermediate frequency data of radar signal is taken as the input of the network to realize the individual identification of radar emitter based on deep learning.The main work of this paper is as follows:(1)This paper introduces the structure principle of the radar transmitter,analyzes the reasons for unintentional modulation in the radar radiation source signal,focus on the analysis of phase noise,builds the radar radiation source individual signal model based on the chirp signal,and proposes the radar radiation source individual identification technology scheme,which provides the basis and support for the research in subsequent chapters.(2)This paper analyzes the reasons for the fingerprint characteristics of radar radiation sources,and extracts different fingerprint characteristics from four aspects of time domain pulse envelope,high-order statistics bisspectrum,frequency domain wavelet transform and transform domain VMD decomposition,and obtains multidimensional feature vectors j ointly.(3)This paper realizes the individual identification of radar emitter based on multidimensional feature matrix.It uses K-nearest neighbor algorithm and support vector machine algorithm for individual identification.It proposed to use logistic regression and gradient boosting decision tree are as classifiers for individual identification respectively,and their feasibility is verified by simulation experiments.By comparison,it is concluded that the gradient boosting decision tree algorithm has a better effect on the individual identification of radar radiation sources based on fingerprint features.When SNR≥15dB,the overall recognition accuracy rate is over 97.5%.(4)This paper proposed to combine the stacked autoencoder and the one-dimensional convolutional neural network with the gradient boosting decision tree respectively to construct a deep learning network,and then uses the network to directly learn and train the intermediate frequency data of the radar radiation source signal,and finally uses the trained network to perform individual identification.The experimental results show that the recognition effect based on the deep learning algorithm is significantly better than the individual recognition of radar radiation sources based on feature extraction.Among them,the one-dimensional convolutional neural network has better effect.And when SNR≥5dB,the recognition accuracy rate reaches more than 99.5%.