Deep Learning Enhanced Code Index Modulation System Empowered by Reconfigurable Intelligent Surface Technology

Abstract

In this study, a new system model (DNN/ML-RIS-CIM) is proposed that predicts code indices with artificial intelligence/machine learning (AI/ML) methods in a reconfigurable intelligent surface (RIS)-assisted code index modulation (CIM) system. Estimating the code indices with conventional detectors can lead to performance limitations. Therefore, instead of the code-index estimation strategy employed in the conventional CIM approach, AI- and ML-based estimators are adopted, enabling the spreading code indices selected at the transmitter side to be identified at the receiver side with higher accuracy. The proposed DNN/ML-CIM-RIS model (using a two-hidden-layer MLP for code-index estimation) captures the nonlinear patterns in despreaded chip vectors and achieves an SNR gain compared with the maximum likelihood detector (MLD).