Attention-based Hybrid CNN-LSTM and Spectral Data Augmentation for COVID-19 Diagnosis from Cough Sound
This repository contains the used source code of the experiments which led to the results presented in the paper Attention-based Hybrid CNN-LSTM and Spectral Data Augmentation for COVID-19 Diagnosis from Cough Sound
We used COUGHVID dataset. See in Zenodo.
The augmented version will be available soon here.
The silence removed version of the dataset will be available soon here.
- utils.py: Provides some helpful functions (progressbar, plotROCCurve, plotCurves, plotConfusionMatrix,...)
- attention_layer.py: Attention layer class
- pitch_shift.py: Run this script to create the signal-augmented version of the dataset
- spec_augment.py: Run this script to create the spectral-augmented version of the dataset (apply SpecAugment technique)
- cnn_baseline.py: Run this script to start CNN model training
- lstm_baseline.py: Run this script to start LSTM model training
- cnn_lstm_baseline.py: Run this script to start hybrid CNN-LSTM model training
- attention_cnn_lstm.py: Run this script to start Attention-based hybrid CNN-LSTM model training
Note: You should change the relative paths in the scripts
Please cite our paper if you find this repository useful.
@article{Hamdi2022,
author = {Hamdi, Skander and Oussalah, Mourad and Moussaoui, Abdelouahab and Saidi, Mohamed},
doi = {10.1007/s10844-022-00707-7},
issn = {1573-7675},
journal = {Journal of Intelligent Information Systems},
title = {{Attention-based hybrid CNN-LSTM and spectral data augmentation for COVID-19 diagnosis from cough sound}},
url = {https://doi.org/10.1007/s10844-022-00707-7},
year = {2022}
}
Feel free to text me in skander.hamdi@univ-setif.dz for any questions or issues in the repository.