Deep Learning-Based Joint Modulation and Coding Scheme Recognition for 5G New Radio Protocols

Abstract

Blind detection of signals is a crucial technique in the 5G/B5G wireless communication systems, especially for the cognitive spectrum radio network, where the parameters of the transmit signals working on the free spectrum can not be known by the receiver. Following the 5G New Radio (NR) protocols, we propose a joint modulation and coding scheme (M-CS) recognition framework based on the supervised learning architecture and the given candidate set of the LDPC encoder. Specifically, the framework is composed of two cascaded modules. Firstly, the type of digital modulation according to the SG NR protocols is recognized blindly based on the proposed Res-Inception convolutional neural network (RICNN). Then, the low-density parity check (LDPC) coding scheme implemented under various bitrates is identified by exhaustively searching the validation candidate to maximize the corresponding average log-likelihood ratio (ALLR). Numerical results show the effectiveness of our proposed blind recognition framework, especially for the practical 5G NR protocols. Moreover, it is demonstrated that our proposed method can guarantee the robustness of the recognition under various channel fading model scenarios.