DroneAudioset: An Audio Dataset for Drone-based Search and Rescue

2025-10-17

AI Summary

Drone-based search and rescue (SAR) faces challenges including strong self-noise, scarcity of real-world audio data, and lack of standardized auditory evaluation protocols. Method: We introduce the first open-source, MIT-licensed drone auditory dataset tailored to realistic SAR scenarios, comprising 23.5 hours of multi-source, in-flight audio recordings across diverse UAV platforms, throttle levels, microphone configurations, and environmental conditions—spanning a wide SNR range—and annotated with fine-grained human speech labels and acoustic feature tags. Contribution/Results: Our dataset uniquely captures natural acoustic interactions between human vocalizations and drone noise under actual flight conditions—overcoming limitations of synthetic data distortion and insufficient diversity in prior datasets. It significantly enhances training robustness and evaluation fidelity of human presence detection models under extreme noise, providing a critical empirical foundation and standardized benchmark for developing drone auditory perception algorithms and hardware–algorithm co-design.

Technology Category

• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Search and Optimization: Sampling/Simulation-based Search
• Humans and AI: Crowd Sourcing and Human Computation

Application Category

• Search and Retrieval-Augmented AI: Personalized, context-aware and across-device search
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

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. Existing datasets are either limited in diversity or synthetic, lacking real acoustic interactions, and there are no standardized setups for drone audition. To this end, we present DroneAudioset (The dataset is publicly available at https://huggingface.co/datasets/ahlab-drone-project/DroneAudioSet/ under the MIT license), a comprehensive drone audition dataset featuring 23.5 hours of annotated recordings, covering a wide range of signal-to-noise ratios (SNRs) from -57.2 dB to -2.5 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 dataset is an important step towards enabling design and deployment of drone-audition systems.

Problem

• Dataset addresses drone ego-noise masking human sounds
• Standardized drone audition setups are currently lacking
• Existing audio datasets lack real acoustic diversity

Innovation

• Dataset for drone audition with real acoustic recordings
• Covers diverse SNRs, drone types, and environments
• Enables noise suppression and human-presence detection methods