Automatic Modulation Detection in Non-orthogonal Multiple Access Systems

Increasing demand and spectrum limitations pose ongoing challenges in telecommunications that constantly engage researchers in this field. Recently, various wireless systems have been developed, including Non-Orthogonal Multiple Access (NOMA), which is proposed for 5G and 6G networks. NOMA improves capacity while maintaining spectral resources by allowing the transmitter to send predefined user symbols with user-specific power over the same resources. The conventional NOMA receiver, Serial Interference Cancellation (SIC), requires knowledge of the modulation types to separate user signals. To improve spectral efficiency and reduce signaling overhead issues, an Automatic Modulation detection algorithm has been proposed. The algorithm enables the receiver to automatically identify the modulation type, eliminating the need for the transmitter to provide this information. This paper presents a novel automatic modulation detection algorithm for a two-user NOMA system that uses feature extraction and deep learning to achieve high detection rates while maintaining manageable computational complexity. The proposed algorithm outperforms previous methods, achieving 93% detection rate at 5 dB signal-to-noise ratio, compared to 78.36% reported in previous studies.