Deep Learning for Radio Frequency Automatic Target Recognition (ATR)

Radio frequency data has long been exploited using Machine Learning (ML) methods, while recent advances in Deep Learning (DL) have grown in popularity with advances in computation, data availability, and various applications. There is a need to understand the merits of DL as applied to radio frequency data for automatic target recognition (ATR). DL-based RF ATR will be beneficial both for object/system recognition for surveillance, self-driving cars, and communications. The focus of this lecture will be recent research results, technical challenges, and directions of DL-based RF ATR. First, an overview past RF ATR research using the MSTAR (Moving and Stationary Target Acquisition and Recognition) data from ML to DL is presented. Next, RF passive exploitation methods will be shown along with connections to information fusion. Finally, the lecture will conclude with future directions of DL-ATR focusing on transparency, interpretability, and explainability as well as contemporary methods. 

U. Majumder, E. Blasch, D. Garren, Deep Learning for Radar and Communications Automatic Target Recognition, Artech House, 2020.