Semi-Supervised Classification for Intra-Pulse Modulation of Radar Emitter Signals Using Convolutional Neural Network

Intra-pulse modulation classification of radar emitter signals is beneficial in analyzing systems. Recently, convolutional neural networks (CNNs) have been used intra-pulse signals, and the results proved better than traditional methods. However, there a key disadvantage these CNN-based methods: CNN requires enough labeled samples. Labeling modulations signal samples tremendous amount prior knowledge human resources. In many circumstances, are quite limited compared with unlabeled samples, which means that will be semi-supervised. this paper, we propose method could adapt approach to case where very number large provided, classify signals. The based on one-dimensional uses pseudo labels self-paced data augmentation, improve accuracy classification. Extensive experiments show our proposed can performance semi-supervised situations.