Exploiting One-Dimensional Convolutional Neural Networks for Joint Channel Estimation and Signal Detection in Non-Orthogonal Multiple Access Systems

Abstract

Non-Orthogonal Multiple Access (NOMA) is a promising technology for the fifth and future generations of wireless communication networks, which increases spectral efficiency and reduces latency. However, NOMA performance can be affected by imperfect successive interference cancellation (SIC). Deep learning techniques have been proposed to aid in signal detection and channel estimation in NOMA systems. In this study, we propose a new approach using one-dimensional convolutional neural networks (1D CNN) to address the limitations of current deep learning methods. Unlike other deep learning methods that rely on time dependencies for data classification, 1D CNN uses a 1-dimensional convolution layer for feature extraction, resulting in high reliability. Simulation results demonstrate that our proposed method outperforms existing deep learning techniques in terms of sample error rate (SER) by 7dB. Moreover, reducing the cyclic prefix (CP) parameter increases inter-sample interference (ISI), but our method still achieves a 6 dB improvement over approaches in [11,13] and traditional channel estimation techniques like maximum likelihood (ML) at low signal-to-noise ratios (SNR).