Self-Supervised RF Signal Representation Learning for NextG Signal Classification With Deep Learning

Deep learning (DL) finds rich applications in the wireless domain to improve spectrum awareness. Typically, DL models are either randomly initialized following a statistical distribution or pretrained on tasks from other domains form of transfer without accounting for unique characteristics signals. Self-supervised (SSL) enables useful representations Radio Frequency (RF) signals themselves even when only limited training data samples with labels available. We present self-supervised RF signal representation method and apply it automatic modulation recognition (AMR) task by specifically formulating set transformations capture characteristics. show that sample efficiency (the number labeled needed achieve certain performance) AMR can be significantly increased (almost an order magnitude) SSL. This translates substantial time cost savings. Furthermore, SSL increases model accuracy compared state-of-the-art methods maintains high is