predicting_alarm_audibility
This Python code is derived from the research article "Evaluating and Predicting the Audibility of Acoustic Alarms in the Workplace Using Experimental Methods and Deep Learning" published in the journal Applied Acoustics. It provides a framework for predicting the audibility of acoustic alarms in noise, and includes scripts to reproduce the results
Science Score: 49.0%
This score indicates how likely this project is to be science-related based on various indicators:
-
○CITATION.cff file
-
✓codemeta.json file
Found codemeta.json file -
✓.zenodo.json file
Found .zenodo.json file -
✓DOI references
Found 7 DOI reference(s) in README -
✓Academic publication links
Links to: sciencedirect.com, zenodo.org -
○Academic email domains
-
○Institutional organization owner
-
○JOSS paper metadata
-
○Scientific vocabulary similarity
Low similarity (11.2%) to scientific vocabulary
Keywords
Repository
This Python code is derived from the research article "Evaluating and Predicting the Audibility of Acoustic Alarms in the Workplace Using Experimental Methods and Deep Learning" published in the journal Applied Acoustics. It provides a framework for predicting the audibility of acoustic alarms in noise, and includes scripts to reproduce the results
Basic Info
- Host: GitHub
- Owner: INRS-France
- License: bsd-3-clause
- Language: Python
- Default Branch: main
- Homepage: https://doi.org/10.1016/j.apacoust.2024.109955
- Size: 1.72 MB
Statistics
- Stars: 1
- Watchers: 1
- Forks: 0
- Open Issues: 0
- Releases: 0
Topics
Metadata Files
README.md
Predicting Alarm Audibility Using Deep Learning
This code is part of an academic publication, and complements the research paper:
F. Effa, J.-P. Arz, R. Serizel and N. Grimault. Evaluating and predicting the audibility of acoustic alarms in the workplace using experimental methods and deep learning, Applied Acoustics, Volume 219, 2024.
Citation
In case you use the code or dataset, please consider citing our paper as: ```bibtex @article{EFFA2024109955, title = {Evaluating and predicting the audibility of acoustic alarms in the workplace using experimental methods and deep learning}, journal = {Applied Acoustics}, volume = {219}, pages = {109955}, year = {2024}, issn = {0003-682X}, doi = {https://doi.org/10.1016/j.apacoust.2024.109955}, url = {https://www.sciencedirect.com/science/article/pii/S0003682X24001063}, author = {F. Effa and J.-P. Arz and R. Serizel and N. Grimault}, keywords = {Psychoacoustics, Alarms, Audibility, Occupational noise, Convolutional Neural Network, Dataset}, }
```
Dataset
A dataset has been collected for the purposes of the study. It is publicly accessible on Zenodo and has to be downloaded to run the programs contained in the present repository.
How to run the code
1) Download the dataset.
2) Unzip data.zip, features.zip, and trained_models.zip under their corresponding sub-folders within the application/ folder.
3) Enter the application/ folder : cd application.
Human baseline performance
python compute_human_performance.py
```python
computehumanperformance.py
Development data - Majority Voting AUROC: 87.68 ± 1.71, F1: 87.70 ± 1.79 Evaluation data - Majority Voting AUROC: 84.53 ± 1.79, F1: 82.78 ± 2.82 Evaluation data - Average Psychometric Function AUROC: 97.01 ± 0.49, F1: 83.48 ± 2.93 ```
Model performance over 10 runs
python compute_trained_models_performance.py
```python
computetrainedmodels_performance.py
AUROC: 85.84 ± 0.75 F1: 76.03 ± 0.57 ```
Audibility criterion: effect on performance of the evaluation label binarization threshold
python binarization_threshold.py computes model and human baseline performance in terms of precision, recall and F1-score for evaluation label binarization thresholds varying
between 0.5 and 1, similar to what was done in the paper.
Continuous model output and human psychometric function
python continuous_output.py Computes the average human baseline psychometric curve and model continuous output values over all the evaluation clips.
Model training
python main_train.py provides an example of code that could be run to train models similarly to what was done in the paper.
Owner
- Name: INRS
- Login: INRS-France
- Kind: organization
- Email: github@inrs.fr
- Location: France
- Website: https://www.inrs.fr
- Repositories: 1
- Profile: https://github.com/INRS-France
Institut National de Recherche et de Sécurité pour la prévention des accidents du travail et des maladies professionnelles
CodeMeta (codemeta.json)
{
"@context": "https://doi.org/10.5063/schema/codemeta-2.0",
"type": "SoftwareSourceCode",
"applicationCategory": "Acoustique",
"auteur": [
{
"id": "_:author_1",
"type": "Person",
"affiliation": {
"type": "Organization",
"name": "INRS - Institut National de Recherche et de Scurit, Rue du Morvan, F-54500 Vanduvre-ls-Nancy, France"
},
"email": "francois.effa@gadz.org",
"familyName": "EFFA",
"givenName": "Franois"
}
],
"codeRepository": "https:/ /github.com/INRS-France/predicting_alarm_audibility",
"contributor": [
{
"id": "_:contributor_1",
"type": "Personne",
"affiliation": {
"type": "Organisme",
"nom": "INRS, Institut National de Recherche et de Scurit, Rue du Morvan, F-54500 Vanduvre-ls-Nancy, France"
},
"email": "jean-pierre.arz@inrs.fr",
"familyName": "ARZ",
" gaveName": "Jean-Pierre"
},
{
"id": "_:contributor_2",
"type": "Personne",
"affiliation": {
"type": "Organisation",
"name": "Universit de Lorraine , CNRS, Inria, Loria, Campus Scientifique, 615 Rue du Jardin-Botanique, F-54506 Vanduvre-ls-Nancy, France"
},
"email": "romain.serizel@loria.fr",
"familyName": "Serizel",
"givenName": "Romain"
}
],
"dateCreated": "2024-05-30",
"datePublished": "2024-05-30",
"description": "Ce code Python est driv de l'article de recherche \"Evaluating and Predicting the Audibility of Acoustic Alarms in the Workplace Using Experimental Methods and Deep Learning\" publi dans la revue Applied Acoustics. Il fournit un cadre pour prdire l'audibilit des alarmes acoustiques dans le bruit et inclut des scripts pour reproduire les rsultats.\n\n\"L'tude prsente galement une technique innovante utilisant un modle de rseau neuronal convolutionnel pour prdire l'audibilit des alarmes dans le bruit. Au-del des critres arbitraires gnriques, cette approche base sur les donnes exploite les connaissances issues d'exemples annots perceptuellement provenant de notre ensemble de donnes contribu. L'valuation des donnes exprimentales et une analyse plus approfondie des rsultats du modle dmontrent un alignement solide des prdictions du modle avec la perception humaine.\"",
"identifier": "https://doi.org/10.1016/j.apacoust.2024.109955",
"keywords": [
"dataset",
"convolutional-neural-network",
"psychoacoustics",
"alarms audibility",
"occupational-noise"
],
"license": "https://spdx.org/licenses/Clause BSD-3",
"name": "Predicting Alarm Audibility",
"programmingLanguage": "Python",
"softwareRequirements": [
"Python 3",
"h5p 3.8.0",
"librosa 0.10.2",
"matplotlib 3.6.0",
"numpy 1.23.5",
"pandas 2.2.2",
"scikit_learn 1.1.0",
"scipy 1.13.1",
"seaborn 0.13.2",
"fichier son 0.12.1",
"torch 2.2.0",
"tqdm 4.66.2"
],
"version": "1.0.0",
"developmentStatus": "active",
"issueTracker": "https://github.com/INRS-France/predicting_alarm_audibility/issues",
"referencePublication": "https://doi.org/10.1016/j.apacoust.2024.109955"
}GitHub Events
Total
- Push event: 4
Last Year
- Push event: 4
Dependencies
- h5py ==3.8.0
- librosa ==0.10.2
- matplotlib ==3.6.0
- numpy ==1.23.5
- pandas ==2.2.2
- scikit_learn ==1.1.0
- scipy ==1.13.1
- seaborn ==0.13.2
- soundfile ==0.12.1
- torch ==2.2.0
- tqdm ==4.66.2