Enhanced UAV-type recognition method using improved EMD and RF fingerprint

Abstract The widespread use of unmanned aerial vehicles (UAVs) has introduced significant security challenges, including unauthorised intrusions and privacy breaches. Conventional identification methods, such as radar and visual surveillance, often struggle to accurately distinguish between different UAV types or individual devices, limiting their effectiveness in countering these threats. To address these concerns, an enhanced empirical mode decomposition method has been developed. This method employs the Welch algorithm to extract power spectral density (PSD) features, which are then used to form the radio frequency fingerprint (RFF). Finally, a statistical classifier is employed to recognise and classify UAVs. The results of experimental studies indicate that this technique yields an average precision of 97% in binary classification (drone vs. no drone) and 84.7% in quaternary classifications (identifying four drone models). A comprehensive performance evaluation, conducted using confusion matrices and receiver operating characteristic (ROC) curves, demonstrates that the classifier performs well across all categories, with low misclassification rates.