This project implements the baseline and benchmarks on Sound-Dr Dataset.

As the burden of respiratory diseases continues to fall on society worldwide, this paper proposes a high-quality and reliable dataset of human sounds for studying respiratory illnesses, including pneumonia and COVID-19. It consists of coughing, mouth breathing, and nose breathing sounds together with metadata on related clinical characteristics. We also develop a proof-of-concept system for establishing baselines and benchmarking against multiple datasets, such as Coswara and COUGHVID. Our comprehensive experiments show that the Sound-Dr dataset has richer features, better performance, and is more robust to dataset shifts in various machine learning tasks. It is promising for a wide range of real-time applications on mobile devices. The proposed dataset and system will serve as practical tools to support healthcare professionals in diagnosing respiratory disorders.

Link: Detail here

This repository contains the official implementation (in Tensorflow+Keras) of the Sound-Dr: Reliable Sound Dataset and Baseline Artificial Intelligence System for Respiratory Illnesses and keep some results in paper.

Our main source code was written and ran on python and JupyterLab in the following directory:

• Baseline System

  • Edit config.py for settings, parameters
  • Train a baseline with command : python3 main.py

• Some results in paper. We do not clean up them and keep some cache files(Feature, fold split csv) because we want to keep original results.

• Overall system. Overall system, include Unsupervised(Isolation Forest, XGBOD) at last this notebook.

  • Chooose dataset dataset_type='SoundDr'
  • chooose pretrain model to extract Feature PRETRAIN="FRILL"
  • Chooose model to classify Classifier="SVM"
  • Set seed seed=2022"

Please, install the following packages

• numpy
• tqdm
• pandas
• zipfile
• pickle
• h5py
• joblib
• librosa
• opensmile
• xgboost
• tensorflow_hub
• scipy
• sklearn

@inproceedings{Truong:2023,
    title={Sound-Dr: Reliable Sound Dataset and Baseline Artificial Intelligence System for Respiratory Illnesses},
    author={Truong, Hoang Van and Quang, Nguyen Huu and Cuong, Nguyen Quoc and Phong, Nguyen Xuan and Hoang, D. Nguyen},
    title={4th Asia Pacific Conference of the Prognostics and Health Management Society (PHMAP)},
    year={2023},
}

This project is based on the following implementations:

• https://github.com/google-research/google-research/tree/master/non_semantic_speech_benchmark
• https://github.com/audeering/opensmile
• https://github.com/DeepSpectrum/DeepSpectrum
• https://github.com/steverab/failing-loudly