Deep learning assisted LDPC decoding for 5G IoT networks in fading environments.

Area of Science:

• Wireless Communication Engineering
• Signal Processing
• Machine Learning for Communications

Background:

• 5G networks and the Internet of Things (IoT) enable advanced applications but face performance challenges.
• Low-Density Parity-Check (LDPC) codes in 5G are sensitive to colored noise and fading channels.
• Colored noise correlation complicates decoding, especially in Rayleigh, Rician, and Nakagami-m fading environments.

Purpose of the Study:

• To enhance the efficiency of LDPC decoding in 5G-enabled IoT networks.
• To address the performance degradation caused by colored noise in fading channels.
• To propose a novel hybrid decoding approach combining deep learning and iterative algorithms.

Main Methods:

• Developed a hybrid decoding architecture integrating the Iterative Offset Min-Sum (OMS) algorithm with a Convolutional Neural Network (CNN).
• Utilized CNN for accurate noise estimation and mitigation in fading channels.
• Employed the OMS algorithm to refine iterative decoding steps and correct noise overestimation.

Main Results:

• The OMS-CNN decoder demonstrated substantial performance improvements over traditional methods.
• Achieved a 2.7 dB enhancement at a specific bit error rate (BER) across diverse fading channels.
• Validated decoder robustness in Rayleigh, Rician, and Nakagami-m fading environments.

Conclusions:

• The proposed OMS-CNN hybrid decoder effectively mitigates colored noise in 5G IoT networks.
• Deep learning integration significantly boosts LDPC decoding performance under fading conditions.
• The approach offers a robust solution for reliable communication in challenging wireless environments.