Unmanned Aerial Vehicles (UAVs) or drones are increasingly used in search and rescue missions to detect human presence. Existing systems primarily leverage vision-based methods, which are prone to fail under low-visibility or occlusion. Drone-based audio perception offers promise but suffers from extreme ego-noise that masks sounds indicating human presence.

In this project, we develop DroneAudioset (The dataset is publicly available at https://huggingface.co/datasets/ahlab-drone-project/DroneAudioSet/ under the MIT license), a comprehensive and largest drone audition dataset featuring 23.5 hours of annotated recordings, covering a wide range of signal-to-noise ratios (SNRs) from -60 dB to 0 dB, across various drone types, throttles, microphone configurations, as well as environments. The dataset enables development and systematic evaluation of noise suppression and classification methods for human-presence detection under challenging conditions, while also informing practical design considerations for drone audition systems, such as microphone placement trade-offs, and development of drone noise-aware audio processing.
This project aims towards enabling the design and deployment of drone-audition systems, and studying human operator and drone collaboration through audio modality for a swarm of drones.