Implementation of the deep learning method for signal detection in massive-MIMO-NOMA systems

Area of Science:

• Optical communications
• Signal processing
• Machine learning

Background:

• Optical nonorthogonal multiple access (O-NOMA) systems enhance spectrum efficiency by allowing shared time-frequency resources.
• Conventional detection methods struggle with complex interference and variable channel conditions in O-NOMA systems.

Purpose of the Study:

• To explore the application of deep learning methods (DLM) for signal detection in massive multiple-input multiple-output (M-MIMO) O-NOMA systems.
• To evaluate the performance of DLM detection against traditional methods.

Main Methods:

• Deep neural networks are designed to process signals, power allocation coefficients, and auxiliary information for symbol decoding.
• Network parameters are iteratively updated using gradient descent optimization with a diverse dataset.
• Comparative analysis of 16x16, 32x32, and 64x64 M-MIMO-NOMA models based on bit error rate (BER), complexity, and power spectral density (PSD).

Main Results:

• DLM algorithms achieve a bit error rate (BER) of 10^-3 at 4.1 dB.
• Excellent power spectral density (PSD) performance of -2500 is demonstrated.
• The proposed DLM algorithms outperform traditional methods with low complexity.

Conclusions:

• Deep learning methods offer a robust solution for signal detection challenges in M-MIMO O-NOMA systems.
• DLM provides significant improvements in BER and PSD performance compared to conventional techniques.
• Careful model design, substantial data, and computational resources are crucial for optimal DLM performance.