Towards Single-Component and Dual-Component Radar Emitter Signal Intra-Pulse Modulation Classification Based on Convolutional Neural Network and Transformer

Abstract

In the modern electromagnetic environment, the intra-pulse modulations of radar emitter signals have become more complex. Except for the single-component radar signals, dual-component radar signals have been widely used in the current radar systems. In order to make the radar system have the ability to classify single-component and dual-component intra-pulse modulation at the same period of time accurately, in this paper, we propose a multi-label learning method based on a convolutional neural network and transformer. Firstly, the original single channel sampled sequences are padded with zeros to the same length. Then the padded sequences are converted to frequency-domain sequences that only contain the amplitude information. After that, data normalization is employed to decrease the influence of amplitude. After radar signals preprocessing, a designed model which combines a convolutional neural network and transformer is used to accomplish multi-label classification. The extensive experiments indicate that the proposed method consumes lower computation resources and has higher accuracy than other baseline methods in classifying eight types of single and thirty-six types of dual-component intra-pulse modulation, where the overall accuracy and weighted accuracy are beyond 90%.