Deep Learning Noncoherent UWB Receiver Design

Abstract

Ultrawideband (UWB) is a promising technology for positioning and wireless communications in Internet-of-things (IoT) applications. However, UWB system's performance is limited by multiple interferences for low complexity noncoherent signal detection methods. Further, deep learning (DL)-based solutions have been envisioned for wireless communications in inaccurate system modeling scenarios. In this letter, we propose a deep learning noncoherent (DLN) UWB receiver to overcome the effect of various interferences such as multiuser interference (MUI), narrowband interference (NBI), and intersymbol-interference (ISI). The DLN is trained offline using the UWB channel statistics, and then, it is used online for data symbol detection. The proposed DLN efficiently learns a nonlinear relationship between input and output in an interference scenario and gives highly accurate data symbol detection, even for training data obtained in a very short period. Numerical results clearly show the proposed DLN UWB receiver's superiority, especially MUI, NBI, and ISI scenarios, over the conventional noncoherent detection method.