Specific emitter identification based on CNN

Abstract

Specific emitter identification is one of the important technologies of electronic countermeasures. The convolutional neural network (CNN) has a strong ability to process image classification. To solve the problem that the bispectral feature of the radiation source does not perform well in the specific emitter identification, this paper proposes a method of combining the two-dimensional bispectrum of the radiation source and CNN. In order to test the performance of the method, this article designed a semi-physical system to collect the radiation source signal. It is verified by experiments that this method has a 92.4% correct rate for identifying different radiation source signals. Compared with other algorithms that combine bispectral features, the proposed method has advantages in accuracy and complexity.