Instructions
- read about human cough types - dry and wet, volunatry/ involuntary, etc. (reference: https://ieeexplore.ieee.org/document/9175986). Why cough is an important biomarker for respiratory disease? Do cough frequency vary with disease severity?
- Select a sound classification dataset containing cough/ lungs/ heart sound e.g. ESC-50 dataset https://github.com/karolpiczak/ESC-50
- make a new dataset with human cough sounds in one class and 4 other types of sounds e.g. sneezing, door knocking, raining, dog barking etc. in another class
- listen to the sounds and visualize some instances/ spectogram using the software Audacity https://www.audacityteam.org/
- label the target class audio records as cough and the other class audio records as not cough (reference: https://www.researchgate.net/publication/336011335_Efficient_Online_Cough_Detection_with_a_Minimal_Feature_Set_Using_Smartphones_for_Automated_Assessment_of_Pulmonary_Patients
- split your dataset into train and test set
- develop a cough classification model using the training set and evaluate the model performance on test set
- Retrain another model using the same dataset with Teachable Machine Platform
- Export the model and evaluate it in python on the test data
- compare your model performance with that of teachable machine
- push your code to project-4 branch of the course Github and make a pull request.