Enhancing Modulation Classification via Diffusion Transformers for Drone Video Signal Processing

Abstract

Reliable drone video signal processing depends on precise identification of modulation type to ensure effective demodulation. Automatic modulation classification (AMC) plays a key role in this process by extracting meaningful features from complex I/Q data. Although deep learning-based approaches have advanced AMC, two challenges still remain: (i) limited support for drone-relevant modulation types and (ii) the need for stable, high-quality generative models for robust data augmentation. This letter proposes the adoption of diffusion transformers (DiT), which capture intricate signal characteristics in diverse drone communication scenarios, including long-range communications, mobile drone networks, and high data rate video transmission. Experimental results demonstrate that DiT improves both the accuracy and robustness of AMC in drone video signal processing scenarios.