Robust Modulation Classification in the Presence of Realistic Propagation in Rural Environment

Automatic Modulation Classification (AMC) is crucial for optimizing spectrum resources in contemporary communication systems by determining the modulation type of received signals without prior information. Deep learning has demonstrated the potential to improve classification accuracy. However, current synthetic datasets lack the consideration of important large-scale fading channel conditions such as shadowing effects, which restrict the accuracy and robustness of modulation classification models. This study aims to overcome this constraint by proposing a method to integrate shadowing effects into the most common synthetic datasets. ITU 1546 standard is used to create an algorithm that models the shadowing effect and improves the generality conditions of modulation classification synthetic datasets. The performance of the proposed model is evaluated through extensive simulations, and compared to prior studies, using CNN and ResNet models, on RML 2016 and RML 2018 datasets. Our research emphasizes the significance of taking shadowing effects into account in AMC and confirms the efficacy of our method in enhancing classification resilience in realistic communication channels.