Feature Extracted-Based Method Against Radar Interference via Deep Learning

The traditional nonparametric / parametric interference suppression methods have some problems, such as weak adaptive ability and the need for interference detection. how to apply the data-driven advantages of deep learning technology to radar interference suppression and realize intelligent interference suppression is a very meaningful and challenging research topic. Considering the anomaly of interference at the feature level of neural networks, a lightweight interference joint detection and suppression model based on feature space distance measurement is proposed in this paper. It can realize interference detection, time frequency positioning and suppression by limiting the normal radar echo feature vector to shrink within the preset hypersphere and minimizing the hypersphere, the results of simulation data and measured data show that the proposed method has low computational complexity and can achieve synchronous interference detection and accurate suppression.

I. Introduction

As an active remote sensing system with all-weather and all-weather advantages, Radar plays an irreplaceable role in civilian and military fields such as terrain detection, resource exploration, and battlefield reconnaissance [1]. However, due to the broadband characteristics of synthetic aperture radar detection signals, their operating frequency range is easily affected by other electromagnetic interference [2]. In the context of the increasingly complex electromagnetic environment, the interference methods for synthetic aperture radar research are becoming increasingly diverse, and the forms of interference are becoming more flexible, which seriously restricts the high-resolution imaging effect and seriously reduces the information acquisition ability of synthetic aperture radar. Therefore, interference suppression technology is one of the important directions in the research of synthetic aperture radar technology.