Robust Automatic Modulation Recognition Through Joint Contribution of Hand-Crafted and Contextual Features

Automatic modulation recognition (AMR) has become increasingly important in the field of signal processing, especially with advancements intelligent communication systems. Deep Learning (DL) technologies have been incorporated into AMR and they shown outstanding performances against conventional methods. The robustness DL-based methods under varying noise regimes is one major concerns for widespread utilization this technology. Furthermore, most existing works neglected contributions hand-crafted features (HCFs) boosting classification In order to address aforementioned technical challenges, a novel robust DL-AMR method proposed by leveraging benefits both contextual (CFs) HCFs specific range signal-to-noise ratio (SNR). A feature selection algorithm also search optimal sets reduce dimensions vectors without losing any relevant features. Simulation studies are performed investigate feasibility classifying 11 types schemes. Extensive performance analyses revealed superiority over baseline terms as well excellent capability determining an subset HCFs.