Radar Emitter Recognition Method Based on Bispectrum and Improved Multi-channel Convolutional Neural Network

Abstract

Aiming at the problem that a kind of bispectrum one-dimensional data is extracted as the individual identification feature of radar emitter resulting in the loss of a lot of useful information in a complex electromagnetic environment, we make full use of the characteristic of symmetry of bispectrum structures, fuse the extracted bispectrum structure, and input an improved multi-channel one-dimensional convolutional neural network in parallel in this paper. Softmax classifier is used for radar emitter identification, so as to improve the recognition rate of individual radar emitter. The simulation experiment uses four kinds of homotype radar to test. Experimental results show that compared with the radar emitter identification method based on 1D-CNN for a single bisspectral one-dimensional feature, the proposed method has the ability of deep feature mining, reduces the loss of useful information, and improves the individual recognition rate of radar emitter effectively.