Testing COVID-19 Cough/Breathing Sound Detection
Cough/breathing sound first level screening for COVID-19
About the Project
We are working on developing a first-level screening for COVID-19 that analyzes cough and respiration sound to detect symptoms such as dry cough (present in 60-70% patients) and shortness of breath (~20% patients). It is targeted for the concerned public who seek recommendations for how to take care of themselves but do not have access to doctors/nurses.
The technology itself is a lightweight neural network that can run on a mobile phone or a small edge device. It should run in real-time, and reliably detect the known symptoms of the disease.
We hope that our project would take its place alongside question-based self-checkers, and help by providing an element of automated and individualized analysis.
We have developed the initial model for the neural network, and have it running on the microcontroller Arduino Nano 33 BLE Sense. Based on the limited dataset we have gathered, the model achieves 70% test accuracy.
However, we have much work to do to ensure the model’s usability for actually analyzing symptoms, including gathering more data and testing the technology on those who have COVID-19.
- Help our mission by providing us with a 10-second recording of your cough sound (link for sharing coming soon!) and displayed symptoms
- For developers: Collecting/annotating data, iteratively developing/testing DL model, inferencing on edge device, web app design
- Who is already working on this
- Helpful links
- How to get in touch
- Number of volunteers needed
- Preferred Volunteer location
- Organization status