Radar Compound Jamming Cognition Based on a Deep Object Detection Network

This article proposes a deep-learning-based compound jamming cognition method to recognize, detect individual jamming elements, and estimate key parameters of them. The method first uses a time–frequency distribution (TFD) to characterize jamming in multiple dimensions (time, frequency, and energy) and then applies an object detection network to identify and locate jamming in the time–frequency domain. This article summarizes the types of jamming parameters and gives corresponding methods for estimating parameters. Unlike traditional studies, this article models jamming recognition as an object detection problem and applies a deep learning framework to find solutions. Therefore, the proposed method has better stability and robustness than conventional techniques, which solves the problem of feature selection caused by the lack of mapping relationship between jamming and features. Another advantage over conventional methods is the multijamming detection capability of this algorithm, which provides more information about individual elements of compound jamming. In terms of jamming parameter estimation, the proposed method makes full use of the geometric characteristics of TFDs, so it is more versatile than conventional methods based on analytical analysis. Simulations and experimental data are used to verify the effectiveness of the proposed method.