FECNet: A Hybrid Deep Learning Network for Automatic Modulation Recognition

Automatic Modulation Recognition (AMR) is essential for non-cooperative wireless communication systems, driving the advancement of future wireless networks. However, under low signal-to-noise ratio (SNR) conditions, existing deep learning-based AMR (DL-AMR) networks continue to face challenges in fully exploiting signal features and maintaining robust performance. To address these issues, we propose a hybrid deep learning network named FECNet. The method first enhances model robustness through a data preprocessing module, followed by a feature enhancement module (FEM) that reinforces key signal features. Furthermore, a hierarchical architecture is employed to progressively extract features through Conv Encoder, producing richer and more discriminative representations. Experimental results show that FECNet outperforms the state-of-the-art DL-AMR methods.