Semester of Graduation
Master of Electrical Engineering (MEE)
Electrical and Computer Engineering
The objective of this thesis is to utilize the power of machine learning to develop a neural network-aided receiver in a DVBS-2X satellite communication system to improve the downlink transmission quality in the presence of interference. An emphasis is placed on mitigating the effects caused by non-linear distortions, carrier frequency offset, and additive white Gaussian noise. This thesis proposes a feed-forward neural network with two hidden layers to compensate for the distortions in the received signal. The proposed system model uses 16-APSK modulation scheme. The neural network is tested under varying degrees of non-linear distortion, frequency offsets, and varying levels of noise power.
Digital pre-distortion and frequency offset correction are used to compare neural network performance. The experiments presented in this work are generated using Simulink, MATLAB, and TensorFlow. The results show that the neural network can outperform digital pre-distortion when only non-linear distortions are considered. The neural network can also compensate for carrier frequency offset when the offset is less severe. However, the proposed neural network struggles when severe frequency offsets are considered. Future work will investigate using neural network architectures with memory to better compensate for severe carrier frequency offset and other channel effects like fading.
Cash, Martha E., "Neural Networks for Interference Mitigation in Satellite Communication Systems" (2022). LSU Master's Theses. 5615.
Available for download on Monday, July 07, 2025