Related Experiment Videos
WUSTCA: enhanced UAV RF signal classification with wavelet transform and STCA attention mechanisms
Jiyu Liu1, Zaolin Xia2, Yuhui Chen3
1College of Information Technology, Shangqiu Normal University, Shangqiu, China.
Scientific Reports
|July 1, 2026
Summary
This study introduces a new method for classifying radio frequency (RF) signals from unmanned aerial vehicles (UAVs) and their controllers. The WUSTCA model achieves high accuracy, offering a reliable solution for real-time UAV applications.
Area of Science:
- Signal Processing
- Machine Learning
- Aerospace Engineering
Background:
- Unmanned aerial vehicles (UAVs) are crucial for civilian and commercial use, requiring precise radio frequency (RF) signal classification.
- Existing deep learning methods face challenges with computational complexity, noise sensitivity, and accuracy limitations.
Purpose of the Study:
- To develop a novel and efficient framework for classifying UAV and controller RF signals.
- To overcome the limitations of current deep learning approaches in UAV signal identification.
Main Methods:
- A hybrid approach combining wavelet-based feature extraction with a hierarchical U-Net architecture.
- Integration of split-time cross attention (STCA) and residual connectivity for enhanced performance.
- Utilized the CardRF dataset for model training and evaluation.
Main Results:
- The proposed WUSTCA model achieved high classification accuracy: 96.6% for UAVs and 95.83% for UAV controllers.
- Demonstrated effectiveness in handling noise interference and diverse signal characteristics.
- Validated the model's reliability and efficiency on the CardRF dataset.
Conclusions:
- The WUSTCA framework offers a robust and efficient solution for UAV signal classification.
- This advancement supports real-time applications of UAVs in complex and noisy environments.
- The study highlights the potential of integrating wavelet features with advanced deep learning architectures for signal analysis.
