A small sample unmanned aerial vehicle radio frequency feature enhancement identification method based on local statistical distribution constraint
This method, which uses local statistical distribution constraints to enhance the radio frequency features of drones in small samples, solves the problem of insufficient sample size, generates feature samples consistent with the target category, and improves the stability and accuracy of drone radio frequency identification. It is suitable for small sample scenarios.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- BEIHANG UNIV
- Filing Date
- 2026-05-15
- Publication Date
- 2026-07-17
AI Technical Summary
In existing technologies for UAV radio frequency identification under small sample conditions, insufficient sample quantity leads to overfitting of the identification model, making it difficult to adapt to signal disturbances in unknown scenarios. Furthermore, general data augmentation methods may disrupt the local structural relationships and statistical regularities of radio frequency signals, resulting in a large deviation between the generated samples and the real signals.
A small-sample UAV radio frequency feature enhancement method constrained by local statistical distribution is proposed, including preprocessing, multi-scale local feature block extraction, iterative update and synthetic feature sample generation, to construct an augmented training set, and combined with a highly adaptive recognition model to generate feature samples consistent with the target category.
Under conditions of limited sample size, it generates stable category augmentation results, maintains consistency between the structural features of augmented samples and real samples, adapts to small sample or even single sample scenarios, and improves the stability and accuracy of the recognition model.
Smart Images
Figure CN122412934A_ABST