CNN-Based Signal Detector for IM-OFDMA

Abstract

The recently proposed index modulation-based up-link orthogonal frequency division multiple access (IM-OFDMA) scheme has outperformed the conventional schemes in terms of spectral efficiency and error performance. However, the induced computational complexity at the receiver forms a bottleneck in real-time implementation due to the joint detection of all users. In this paper, based on deep learning principles, a convolutional neural network (CNN)-based signal detector is proposed for data detection in IM-OFDMA systems instead of the optimum Maximum Likelihood (ML) detector. A CNN-based detector is constructed with the created dataset of the IM-OFDMA transmission by offline training. Then, the convolutional neural network (CNN)-based detector is directly applied to the IM-OFMDA communication scheme to detect the transmitted signal by treating the received signal and channel state information (CSI) as inputs. The proposed CNN-based detector is able to reduce the order of the computational complexity from O(n2n) to O(n2) as compared to the ML detector with a slight impact on the error performance.