Summary
International Conference on Emerging Technologies for Communications
2023
Session Number:P1
Session:
Number:P1-16
Neural Networks as Catalysts for Enhanced Lossless Compression in Satellite Communications
Thibault Piana, Matthieu Arzel, Abdeldjalil Aﾃｯssa-El-Bey, Alain Thomas,
pp.-
Publication Date:2023/11/29
Online ISSN:2188-5079
Summary:
Satellite communication systems enter the New Space era with a growing pressure on the ground segment, whose architecture converges to that of mobile networks with Centralized Radio Access Networks (C-RAN). Lossless data compression is a key feature for such systems, for which we propose an alternative strategy to conventional methods, such as Linear Predictive Coding (LPC), by using nonlinear predictors based on neural networks that show substantial performance improvements in high-SNR conditions (compression ratio increased by more than 10 percentage points with some architectures) while preserving signal quality. These neural networks demonstrate robustness by maintaining comparable performance with LPC in low-SNR conditions (Es/N0 < 10dB). We provide a comprehensive comparison between these methods under varying SNR conditions and delve into the influence of network parameters on performance. Our results suggest that neural networks could serve as an effective tool for improving lossless data compression in modern communication systems.