Modulation Recognition using YOLOv5 on the WBSig53 Dataset

Abstract

This paper discusses applying a set of YOLOv5 neural networks to the WBSig53 dataset in order to perform modulation recognition. Identifying modulation schemes is a main step in the development of smart receivers. In this effort, attention is payed to the amount of time needed for training as well as inference speed. Signal detection, modulation family classification, and individual modulation scheme recognition are explored on clean and impaired WBSig53 data.