Treffer: Fusion-Driven Deep Learning for Automatic Modulation Classification Using Constellation Imagery and Fading Features.

Automatic Modulation Classification (AMC) is a key enabler for intelligent and adaptive wireless communication systems, especially in the domains of 5G, 6G and cognitive radio networks. In this research, we propose a novel deep learning-based AMC framework leveraging constellation images (CIs) combined with auxiliary channel fading information. A dataset was generated using Python, covering a broad Signal-to-Noise Ratio (SNR) range from -10 dB to 10 dB, and simulating Rayleigh fading effects with a Beta-distributed fading scale, closely mimicking realistic multipath environments. The proposed model employs a fusion architecture where constellation images are processed through an EfficientNet-B0 convolutional backbone while fading values are separately processed through a lightweight neural network. The extracted image and fading features are concatenated and jointly classified using a fully connected network. The findings in this research establish the viability of using constellation image-based fusion models for real-time AMC applications in software-defined and cognitive radio systems. [ABSTRACT FROM AUTHOR]