Multi-User Joint Detection Using Bi-Directional Deep Neural Network Framework in NOMA-OFDM System.

Area of Science:

• Electrical Engineering
• Computer Science
• Telecommunications

Background:

• Non-orthogonal Multiple Access (NOMA) is crucial for 5G wireless communication.
• Traditional successive interference cancellation (SIC) methods struggle with complex channels and error propagation.

Purpose of the Study:

• To propose a deep learning model for enhanced multiuser uplink channel estimation and signal detection in NOMA systems.
• To overcome the performance limitations of traditional NOMA detection methods.

Main Methods:

• A deep neural network utilizing Bi-directional Long Short-Term Memory (Bi-LSTM) was developed.
• The Bi-LSTM model was trained offline using simulated channel data and deployed online for signal recovery.

Main Results:

• The proposed Bi-LSTM model directly recovers multiuser signals distorted by channel effects.
• Performance was compared against Convolutional Neural Networks (CNN), Minimum Mean Square Error (MMSE), and Least Squares (LS) methods.

Conclusions:

• The Bi-LSTM approach demonstrates significant improvements in symbol-error rate and signal-to-noise ratio.
• This deep learning method is well-suited for 5G wireless communication and future advancements.