Wireless communication are becoming inherently complex with every passing generation under varying mobility and channel imperfections. In 5G NR and Beyond-5G, multi-user Multi-input-multi-output are highly susceptible to both imperfections further constraints conventional model-based receivers in estimation and detection of the received signal. This thesis explores the advent of neural network-based receivers as data-driven alternatives to conventional receivers for single-input single-output system. With the focus on single and multi-carrier systems, this thesis aims to provide a holistic view on the performance of the neural network receivers under varying hyperparameters. To assess its performance, the neural network receivers are benchmark against optimal and suboptimal detectors, including MLSE, MMSE and ZF under various channel configurations. The results highlight that the data driven neural network receivers in both outperforms the suboptimal detectors.
| Date of Award | 2026 |
|---|
| Original language | English (American) |
|---|
| Awarding Institution | |
|---|
| Supervisor | Hans-Georg Brachtendorf (Supervisor) |
|---|
Deep Learning-Based Receiver for Future Communication Systems
Lee, D. T. (Author). 2026
Student thesis: Master's Thesis