Index Modulation Multiple Access via Deep Learning based Detection

Abstract

In this paper, we propose a downlink index modulation multiple access (IM-MA) system using deep learning (DL) based detection. In the proposed IM-MA, a user transmits information by modulating either active antenna indexes or signal constellation points, unlike the conventional IM-MA, where each user sends information using both the antenna indexes and constellation points. Therefore, the proposed IM-MA can accommodate more users in a network. Further, we use the DL-based detector via deep neural network (DNN) models, for each user’s symbol detection to improve the proposed IM-MA system’s performance. The received signal is preprocessed by considering the system’s apriory knowledge before going into the DNNs. DNN models are trained offline via simulated data and then applied for online symbol detection. Simulation results show the effectiveness of DNN detectors in terms of symbol error rate performance over Rayleigh fading channels with a lower runtime and complexity as compared to optimal maximum-likelihood detection.