Deep Learning-Assisted Automatic Modulation Classification in Adaptive FBMC/OQAM Systems

This manuscript presents a novel blind modulation classification method applied to adaptive FBMC/OQAM systems which utilizes a neural network to combine different information types for distinguishing between different modulation formats. In the considered bit-loaded multicarrier transmission system, the receiver classifies the bit allocation table based on the received signal. Since the transmission channels from transmit and receive direction are correlated in time-division duplex systems, important side information about the used modulation schemes on each subcarrier is available, thus enabling a much more reliable classification. A deep learning-based algorithm is investigated numerically that extracts features from preprocessed data such as likelihood and receiver bandwidth efficiency values. Simulation results show a superior performance in typical wireless local area network scenarios compared to existing approaches.