Research On Radar Emitter Recognition Method Based On Machine Learning

As a key part of the radar reconnaissance system,radar emitter recognition plays a crucial role in electronic warfare.Through the characteristics analysis of radar emitter signal,radar type and working mode can be confirmed to further assess its role in warfare,tactical characteristics,and menace grade.With the development of information technology,how to realize the accurate recognition of the radar emitter signal has been a hot research topic under the conditions of the continuous deterioration of the electromagnetic environment and the increasingly complex and variable conditions of the radar modulation method.This thesis describe the research background and development status of radar emitter recognition technology,and the problems and challenges faced in current radar emitter recognition field are pointed out.Then,based on the understanding of radar emitter recognition knowledge,the machine learning methods,such as deep learning and data mining,are applied to the recognition of radar emitter,which enables effective recognition of radar emitter and provides a basis for the identification of radar emitter' working modes.The main contents of this thesis are as follow:1.A radar emitter recognition method based on SVM classifier is studied in this thesis.Firstly,the description of the radar emitter such as the pulse description word and the modulation mode and range of variation of each parameter is studied.Then for the radar emitter recognition method,the multi-classification method of SVM classifier and the selection of the kernel function are explained.Finally,the effects of different multi-classification methods of the SVM classifier and different kernel function on the recognition of radar emitters are experimentally analyzed.2.A radar emitter recognition method based on deep learning is proposed.Combining with the stacked autoencoder and deep belief network,a deep neural network model is constructed for the recognition of radar emitters.The algorithm model,training method and fast learning algorithm are studied.The influence of network parameters on the recognition result is analyzed.The method solves the problem of poor adaptability to cruel electromagnetism environments in common radar emitter recognition methods.The method can also avoid artificial feature extraction of the radar emitter signals,and can automatically extract the deep features of radar emitters in the time domain signal and frequency domain signal,which realize the accurate identification of radar emitter signal.3.Radar emitter working mode recognition method based on data mining is studied in this thesis.Firstly,the special features and parameter scope of common radar working modes are summarized.Then,combined with the association analysis in data mining,the frequent item set of radar emitter signal is excavated through Apriori algorithm,which contributes to the estimate of of radar emitters' working mode.The Apriori algorithm is improved according to the characteristics of the radar emitter signal,which improves the efficiency of the algorithm effectively.