Radar Emitter Signal Identification Based On Deep Learning Technology

Radar Emitter Identification(REI)has been a long-standing topic in military and civil fields.The essence of radar emitter identification is the problem of pattern recognition.Through the realization of important technologies such as the feature extraction,selection and classification of radar radar emitter's characteristic parameters,the relevant weapon type classification,platform and even radar individual identification are completed,thereby clearly knows the battlefield electromagnetic environment situation and provide accurate auxiliary basis for commander's leadership.Firstly,dramatic increase in the electronic countermeasures and the emergence of various radar often reduce the classification performance of traditional recognition algorithms,sometimes even make great mistakes.To this end,this paper proposes a novel radar emitter identification approach,based on the bispectrum of radar emitter,and use the efficient hierarchical extreme learning machine(H-ELM)for feature representation and feature classification.Simulations of six typical radar signals verifies its efficient performance.Secondly,in such high density,complex and variable signal modulation environment,although the existing deep learning algorithm can barely achieve the recognition accuracy of the radar emitter,the convergence of the neural network requires a large amount of training time,and it is difficult to meet the Real-time requirement of future wars.Therefore,combined with the existing multi-layer learning machine algorithm and direction gradient histogram(HOG)algorithm,the feature extraction and classification of radar emitters are carried out,and the effectiveness of this algorithm are verified.Finally,based on the bispectrum local binary pattern(LBP)and bispectral gradient histogram of the radar emitter,the fusion feature is used as the classification feature,and then the extreme learning machine is used for classification.The algorithm is demonstrated by typical radar emitters.Extensive experiment results demonstrate the feasibility of the proposed method.