Deep-learning based joint detection and decoding for non-orthogonal multiple-access systems

• Main Author: Huang, Zemin
• Other Authors: Teh Kah Chan
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2021
• Subjects: Engineering::Electrical and electronic engineering::Wireless communication systems
• Online Access: https://hdl.handle.net/10356/149283
• Institution: Nanyang Technological University

As non-orthogonal multiple access (NOMA) system is gaining its popularity in fifth generation (5G) network and beyond due to its superiority in bandwidth and connectivity, the concerns of drawbacks in NOMA decoding method, successive interference cancellation (SIC), is raised in this report. Moreover, due to the unstable and rapidly changing channel condition, conventional methods in channel estimation such as least square (LS) and minimum mean-square error (MMSE) have fallen short. Therefore, this report presents a novel approach, deep learning (DL), to carry out channel estimation and decoding jointly in a NOMA system. Different from traditional methods, DL acts as a black box that replaces sub-blocks such as slicing, multiplexing and modulation in the traditional methods, and recovers the received signals that have suffered from channel distortion back to the original transmitted signals at one go. Three diverse deep learning networks: long short-term memory (LSTM), convolutional neural network (CNN) and deep neural network (DNN), are designed to analyse the efficiency and performance of DL-based NOMA. The results obtained from the respective neural network models have shown that the proposed DL-based NOMA system could achieve a better performance than conventional ones with maximum likelihood (ML) as the benchmark, along with CNN attaining the best performance in terms of bit-error rate (BER). Through further evaluation, it can be concluded that DL is an effective way of reducing the computational complexity and at the same time enhancing the decoding accuracy of signals in NOMA system.