Deep Learning based Random Access Preamble Detection for 3GPP NB-IoT Systems

Abstract

Narrowband Internet of things (NB-IoT) is a promising cellular IoT standard from the third generation partnership project (3GPP), designed to provide extended coverage and connectivity to a large number of low-cost devices. An NB-IoT user equipment (UE) requiring network access first transmits a preamble on the NB-IoT physical random access channel (NPRACH). The NPRACH receiver at the NB-IoT base station (eNB) detects the preamble, estimates the time-of-arrival (ToA) and calculates the residual carrier frequency offset (RCFO) for each UE. Conventional NPRACH receivers are based on cross-correlation and energy-based detection. In this work, a novel deep learning based NPRACH receiver is developed. It is demonstrated that the proposed deep learning based receiver not only performs nearly as good as the traditional receiver, but is also computationally feasible and scalable.