Automatic Modulation Classification for Aggregated PDSCH in 5G New Radio

Abstract

In this paper, we propose a new automatic modulation classification (AMC) scheme for aggregated physical downlink shared channel (PDSCH) in 5G New Radio (NR). First, we develop a generic convolutional neural network (CNN) structure for classifying modulation types specified in 5G NR which are QPSK, 16QAM, 64QAM, and 256 QAM. The received n aggregated PDSCHs are divided into multiple segments, then the received symbols within same group are softly combined together. After that, a set of combined symbols are input to the CNN-based classifier. Then, the classification results are further fused in three different ways to improve the classification accuracy. From the simulations, it is observed that the classification accuracy can be maximized by adjusting the soft-combining level for the received PDSCHs considering trade-off between signal-to-noise ratio gain and diversity gain.