Resumo

The Internet of Drones (IoD) refers to a robust mobile network with well-defined airways where drones perform interoperable services, enhancing the deployment of Smart Cities. Automatic Drone Identification (ADI) is a protection mechanism to detect and avoid malicious drones, where different techniques have been used, such as acoustic signals. In this field, the sound generated by propellers and motors has particular characteristics, being a potential aspect to explore; however, this investigation is still missing. This study examines the use of rhythm-based descriptors as input features to ADI, based on the hypothesis that the acoustic signal generated by different drones has different rhythmic properties. Aiming to explore and validate our approach, we formulate an ADI methodology using rhythm-based features. We use a freely available drone audio dataset, comparing our results with a baseline study. As a result, our classification model improves 3.47% the baseline binary classification and 2.97% the multiclass classification, reaching accuracy rates of 0.9985 and 0.9591, respectively. Although the improvements are narrow, they point out that acoustic features have a great potential to enhance ADI mainly in dense drone-based environments, where drone identification is an essential task