Audio-based classification of drones – AuDroK
Drones are operated more and more frequently. This has advantages, but it also causes risks and dangers for other aircraft, infrastructures and third parties. To minimize these risks, government-provided legal guidelines are needed, and their implementation requires effective detection and classification technology. A very well-suited and economical technology for this purpose is analyzing drone sounds with the help of machine learning algorithms. The aim of the AuDroK Project is to compile a dataset containing the sounds of different drone models and to develop classifiers that allow the identification of drones based on their sounds. The project will end with a model experiment for assessing the performance of the approaches developed. The project's successful completion will mean a contribution to improving the safety of drone flights and facilitating the integration of drones into our society.