AeroSonicDB 3K: Audio Dataset of Low-Flying Aircraft

AEROSONICDB 3K: AUDIO DATASET OF LOW-FLYING AIRCRAFT

Version 2.0-beta (July 2024)

This repository is currently restricted, however you can gain access by "requesting access" on this page with your Zenodo account, or by emailing Blake Downward at [e-mail], or by submitting the following Google Form.

Created by

Blake Downward

Publication

When using this data in an academic work, please reference the dataset DOI and version. Please also reference the following paper which describes the methodology for collecting the dataset and presents baseline model results.

Downward, B., & Nordby, J. (2023). The AeroSonicDB (YPAD-0523) Dataset for Acoustic Detection and Classification of Aircraft. ArXiv, abs/2311.06368.

Description

AeroSonicDB 3K is a machine labelled audio dataset of over 3,000 low-flying aircraft events. Samples in this dataset were recorded at locations in close proximity to a flight path approaching or departing Adelaide International Airport’s (ICAO code: YPAD) primary runway, 05/23. Recordings are triggered and labelled based on radio (ADS-B) messages received from the aircraft overhead, then human verified and trimmed to the first and final moments the target aircraft is audible in the sample.

Major changes from version 1

This iteration of AeroSonicDB was developed with the aim to improve accessibility and usability for the end-user, streamline the sample collection and annotation procedures, and minimise memory consumption without losing crucial information. A summary of the major changes include:

Sample rate for audio files is now 16,000 Hz (previously 22,050 Hz).

Samples have been trimmed to the audio event onset and offset moments - the entire duration of the clip can be assumed to be of the same class.

“altitude” has been ommited as a sample feature.

“engfamily” and “shortdesc” have benn ommitted as aircraft features.

Samples are split across 5 temporaly disjointed folds. (previously had a sixth fold to act as a hold-out test set)

Introduction of a PDM microphone embedded in a microcontroller for capturing audio (approximately 20% of all samples)

A machine learning model trained on the original dataset was deployed to assist with annotation.

Approximately 60% of newly acquired silence/background samples were annotated by progressive iterations of this model.

582 unique aircraft recorded (up 280 from 302 in version 1)

3,058 recordings of aircraft audio events for a total of 38 hours (up from 625 recordings and 8.9 hours)

32 hours of silence/background audio (up from 3.5 hours)

Audio data

SAMPLE RATE: 16,000 HzDATA TYPE: ‘int16’/‘wav’

Class distribution (hours of audio)

Binary Classes

no aircraft/background (32)

aircraft (38)

Sub-classes

no aircraft/background (32)

Piston aircraft (1.2)

Turboprop aircraft (5.3)

Turbofan aircraft (31)

Rotorcraft/helicopter (0.5)

Data splits/folds

Recordings have been split into 5 folds, giving researchers a common split for cross-validation and ensuring comparable results. Folds are temporaly disjoint to avoid data leakage.

Metadata files

The entire dataset is referenced with labels in the “sample_meta.csv” file. Each row contains a reference to a unique recording, as well as meta information such as; class, duration and fold.

The “aircraft_meta.csv” a file can be used to reference aircraft specific features - such as; manufacturer, engine type, ICAO type designator etc. (see “Columns/Labels” below for all features).

Audio-file naming convention

Audio samples are in WAV format, with some additional metadata stored in the filename.

Basic Convention

“Aircraft ID + Date + Time + Location ID + Microphone ID”

“XXXXXX_YYYY-MM-DD_hh-mm-ss_X_X”

Sample with aircraft

{hex_id} _ {date} _ {time} _ {location_id} _ {microphone_id} . {file_ext}

7C7CD0_2023-05-09_12-42-55_2_1.wav

Sample without aircraft

“Silence” files are denoted with six (6) leading zeros rather than an aircraft hex code. All relevant metadata for “silence” samples are contained in the audio filename, and again in the accompanying “sample_meta.csv”

000000 _ {date} _ {time} _ {location_id} _ {microphone_id} . {file_ext}

000000_2023-05-09_12-30_2_1.wav

Columns/Labels (sample_meta.csv)

filename: The filename of the audio recording

duration: Length of the audio event in seconds.

date: Date of the recording

fold: Digit from 0 to 4 splitting the dataset 5 ways

class: Class label for the recording (eg. 0 = No aircraft/background, 1 = Aircraft audible)

hex_id: Unique ICAO 24-bit address for the aircraft recorded

Columns/Labels (aircraft_meta.csv)

hex_id: Unique ICAO 24-bit address for the aircraft recorded

Manu: Aircraft manufacturer (eg. Boeing, Pilatus, Airbus)

Model: Aircraft model (eg. 737-800, A320-232, DHC-8-315)

engnum: Number of engines

Engmanu: Engine manufacturer (eg. Pratt & Whitney, CFM International, Rolls Royce)

Engtype: Type of engine (eg. Piston, Turboprop, Turbofan, Turboshaft)

Engmodel: Engine model (eg. TRENT XWB, CFM56-7B24E, PT6E-67XP)

Fueltype: Fuel type used in the engine (eg. Gasoline, Kerosene)

Airframe: Describes the mechanical structure of the aircraft (eg. Power Driven Aeroplane, Rotorcraft)

Propmanu: Propeller manufacturer (eg. Hartzell Propellers, Hamilton Standard, “Aircraft Not Fitted With Propeller”)

Propmodel: Propeller model (eg. HC-E5A-3A/NC10245B, 14SF-15, “Not Applicable”)

MTOW: Maximum take off weight (MTOW) in kilograms

ICAOtypedesig: ICAO type designator for make and model of aircraft (eg. PC12, C185, B738)

Conditions of use

Dataset created by Blake Downward.

The AeroSonicDB 3K dataset is offered free of charge for non-commercial use under the Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0) license.

https://creativecommons.org/licenses/by-nc/4.0/

Acknowledgements

Special thanks to Jon Nordby of Soundsensing AS - his contributions were pivotal in maximising the potential of this dataset for open-source release.

Feedback

Please send suggestions, feedback and comments to: Blake Downward: [e-mail]

Change log

2.0-beta: All major changes as described above

1.1.2: Added “environment_mappings_raw.csv”. No change to audio from Version 1.1

1.1.1: Minor change to “sample_meta.csv” - replaced “6” with “test” in the “fold” column

1.1: Replaced truncated aircraft samples with the original full-length files and annotated the beginning and end of each audio event. Added ‘ignore’ statements to aircraft event boundaries in the environmental class mappings file.

1.0: Environmental audio and mappings added

0.3: locations.json file added, README updated

0.2: location information added to README