Deep Learning Based Classification of CP-OFDM RAT-Dependent Signals

Spectrum sharing is one of the key concepts in current Radio Access Technologies (RAT) and specially for 5G. The ability to detect and classify signals present in a specific band is relevant to optimize the use of these shared channels and to avoid interference. One of the most promising tools for this is Deep Learning. This paper explores several Deep Learning architectures to classify different CP-OFDM RAT signals, based on the difference in symbol length and cyclic prefix length parameters of these technologies.