I. Introduction
In Recent years, with the rapid development of deep learning, the convolutional neural network has been widely used for radar signal recognition [1]–[4]. Researchers converted radar signals into time-frequency images through time-frequency analysis and used convolutional neural networks for automatic feature extraction and classification by supervised learning [4]–[9]. The radar signal recognition based on the convolutional neural network significantly improves the algorithm efficiency and recognition accuracy [10], [11]. However, in an open environment, the receiver will intercept novel radar signals. How to recognize novel radar signals with few samples in an open environment is still a challenge [12]–[17]. The subject of few-shot learning is to learn to recognize previously unseen classes with very few annotated examples.