A New Error Detection Method with Neural Networks-based Equalizer and Reject Option

Abstract

In this paper, we propose a method combining equalization and error detection using neural networks with reject option to address challenges of intersymbol interference and noise in communication channels. Initially, we compare our proposal with three conventional neural networks for channel equalization, noting an improvement in performance of 9-12%. Subsequently, we compare our approach with a state-of-the-art algorithm, analyzing bit error rate (BER) curves across various communication channels. Finally, we present the results of our proposal compared to error detection methods, highlighting its ability to detect errors without additional data overhead and its superiority in high-noise environments.