Limitations of Gaussian Mixture Models for Symbol Detection in Harsh Electromagnetic Environments

Abstract

This letter presents a study on the limitations of Gaussian mixture models (GMMs) for symbol detection in digital communication systems. Although GMMs have proven to be adequate for symbol detection at medium signal-to-interference ratios, the results of this study reveal a critical flaw in the conventional assumption that Gaussian models can accurately represent the received symbol distributions under all circumstances in harsh environments. This statistical approach proves to be inadequate at high levels of electromagnetic (EM) interference, resulting in an inability to accurately differentiate between symbols. The methodology is tested in a triple modular redundant system subjected to reverberant EM disturbances, with diverse PAM-4 as an encoding technique.