An anonymized, privacy-preserving companion dataset for studying long-term, distributed dB-only acoustic exposure dynamics in occupied office buildings. It is the public data release accompanying the manuscript "Privacy-Preserving Long-Term Acoustic Exposure Dynamics in an Occupied Multi-Floor Office Building Using Distributed dB-Only IoT Monitoring."
The dataset contains aggregate feature tables, mixed-effects model outputs, cross-deployment validation summaries, a real device-vs-reference-microphone calibration check, and publication figures. It does not contain raw audio, raw high-frequency sound logs, exact calendar dates, clock times, customer names, team labels, original floor labels, or labeled floor-plan images.
- Release Status
- Dataset Snapshot
- Pilot Overview
- What Is Included
- Key Use Cases
- Scientific Boundaries
- Repository Structure
- How To Use
- Citation
- Zenodo Archiving
- License
- Contact
- Current release:
v1.1.0, archived on GitHub and Zenodo - See
RELEASE_NOTES.mdfor the full version history - GitHub repository: https://github.com/Pouya-Mansournia/acoustic-exposure-dynamics-dataset
- Zenodo DOI: 10.5281/zenodo.21503880
- Project type: scholarly dataset and manuscript companion material
| Item | Public description |
|---|---|
| Main deployment | Deployment_A: 56 pseudonymized nodes across five anonymized floors |
| Validation deployment | Deployment_B: 13 pseudonymized nodes on one anonymized floor |
| Underlying records | 185,140,775 in Deployment A; 24,393,582 in Deployment B |
| Signal | Calibrated SPL-like dB values derived from 5 ms RMS windows |
| Calibration check | One in-situ session, one device vs. a UMIK-1 reference microphone: r = 0.87, mean bias +8.64 dB |
| Raw audio | Not stored and not released |
| Public data level | Aggregate features, summaries, model outputs, and figures |
SPL-like means device-calibrated sound-level values derived from deployed firmware. It should not be interpreted as IEC-certified sound-level-meter output.
data/processed/main_deployment/— Deployment A aggregate features, pairwise metrics, mixed-effects outputs, and sensitivity analyses.data/processed/validation_deployment/— Deployment B aggregate validation features.data/processed/cross_deployment/— Cross-deployment replication summaries.data/processed/calibration_validation/— A single continuous, unscripted in-situ co-location session between one Acust device and a calibrated miniDSP UMIK-1 reference microphone, time-aligned to 1-second bins by cross-correlation (r = 0.87, mean bias +8.64 dB, RMSE 10.15 dB; seedocs/privacy_and_limitations.mdfor full scope and caveats).data/metadata/— Public manifest, data dictionary, and pseudonymized sensor keys.figures/manuscript/— Publication-ready, anonymized manuscript figures.docs/— Methodology notes and privacy/limitations documentation.
- Long-term office acoustic exposure analysis.
- Privacy-preserving acoustic monitoring research.
- Operational-regime comparison across workday, closed-day, and activity-window conditions.
- Testing whether average dB hides variability, high-exposure share, and quiet-period differences.
- Studying device-level calibration accuracy against a reference measurement microphone.
- Reproducing the public model, sensitivity, and validation outputs reported in the manuscript.
This dataset supports operational acoustic observability, not standardized room-acoustic characterization or certified sound-level-meter compliance measurement. It cannot support claims about RT60, T20/T30, room impulse responses, absorption coefficients, speech content, sound-source identity, exact occupancy counts, speech transmission index (STI), or physical propagation time between nodes.
Event-level categories (closed days, low-attendance periods, broad remote-work periods) are operational case observations rather than statistically replicated interventions. The calibration-validation data is a single-device, single-session, ambient-stimulus check — real and timestamped, but not a substitute for a controlled, multi-device, multi-level laboratory calibration campaign.
data/
metadata/
public_manifest.json
data_dictionary.md
main_sensor_key_public.csv
validation_sensor_key_public.csv
processed/
main_deployment/
validation_deployment/
cross_deployment/
calibration_validation/
figures/
manuscript/
docs/
methodology_notes.md
privacy_and_limitations.md
- Start with
data/metadata/public_manifest.jsonfor the release inventory and privacy transform. - Use
data/metadata/data_dictionary.mdto interpret public fields, including the calibration-validation columns. - Load the CSV files under
data/processed/for aggregate analysis, model review, or manuscript reproduction. - Read
docs/methodology_notes.mdanddocs/privacy_and_limitations.mdbefore drawing conclusions from any table — they state what each metric does and does not represent. - Use the figures under
figures/manuscript/as public, anonymized visual summaries.
If you use this dataset, cite this repository and the associated manuscript when available:
Mansournia, P. (2026). Acoustic Exposure Dynamics Dataset: Anonymized long-term dB-only office sound-level features for privacy-preserving acoustic monitoring research (v1.1.0). Zenodo. https://doi.org/10.5281/zenodo.21503880
See CITATION.cff for machine-readable citation metadata.
This repository is archived on Zenodo:
- Concept DOI (always resolves to the latest version): 10.5281/zenodo.21503880
- GitHub releases: https://github.com/Pouya-Mansournia/acoustic-exposure-dynamics-dataset/releases
- Zenodo metadata source:
.zenodo.json
The public aggregate dataset and documentation are released under the Creative Commons Attribution 4.0 International License.
Private raw data, raw logs, customer metadata, and any non-public deployment materials are not included in this release and are not licensed by this repository.
For questions about this dataset, the associated manuscript, or requests for additional (privacy-reviewed) analysis code, contact the corresponding author, Pouya Mansournia, or open an issue in this repository.