A Deep Learning Approach for MIMO-NOMA Downlink Signal Detection

Area of Science:

• Wireless Communication
• Signal Processing
• Artificial Intelligence

Background:

• Non-orthogonal Multiple Access (NOMA) is a key technology for 5G systems.
• Successive Interference Cancellation (SIC) is the standard NOMA detection method but suffers from complexity and error propagation.
• Deep Learning (DL) presents a promising alternative for complex signal processing tasks.

Purpose of the Study:

• To propose a novel Deep Learning (DL) method for Non-Orthogonal Multiple Access (NOMA) signal detection.
• To address the limitations of traditional Successive Interference Cancellation (SIC) in NOMA systems.
• To enhance the performance and efficiency of 5G communication systems.

Main Methods:

• Developed a DL-based method for analyzing Channel State Information (CSI) and detecting transmit sequences.
• Integrated channel estimation and signal recovery within a single DL framework.
• Utilized a Neural Network (NN) to find optimal solutions for NOMA signal detection in MIMO-NOMA systems.

Main Results:

• The proposed MIMO-NOMA-DL system demonstrated superior detection performance compared to conventional SIC.
• The DL method effectively mitigated channel impairments and multiuser signal superposition.
• Achieved high-performance and high-efficiency signal detection in simulated NOMA environments.

Conclusions:

• Deep Learning (DL) is a powerful and effective tool for NOMA signal detection.
• The proposed DL approach overcomes the limitations of SIC, offering improved accuracy and efficiency for 5G.
• DL provides a robust solution for complex signal processing challenges in advanced wireless communication systems.