AIRA-UAS: an Evaluation Corpus for Audio Processing in Unmanned Aerial System

This paper presents the first version of the AIRA-UAS corpus. It is a set of recordings produced by the ego-noise of an Unmanned Aerial Vehicle (UAV) performing different aerial maneuvers. We also recorded audios produced by other drones flying near the UAV capturing the audio signals on board. The aim of this corpus is to provide an evaluation mechanism for sound source localization and separation algorithms, where the sound data capture process is carried out on board an UAV. We argue that this corpus will be useful for the development of UAV applications focusing on search & rescue operations as well as for detection of unauthorized drone operation. In addition, we also argue that our corpus may prove useful to assess the impact level at which the noise produced by drones affects the welfare of human beings and wildlife.