When you hear a human sound sample, you begin to infer - do you know the person, what is s/he saying, the personality, health status, etc. Drawn by the information packed nature of sound signals, Project Coswara aims to evaluate effectiveness for COVID-19 diagnosis using sound samples. The idea is to create a huge dataset of breath, cough, and speech sound, drawn from healthy and COVID-19 positive individuals, from around the globe. Subsequently, the dataset will be analysed using signal processing and machine learning techniques for evaluating the effectiveness in automatic detection of respiratory ailments, including COVID-19. We do not know if we will succeed in this but it is worth a try, and to us, now looks the right time to give a shot. We are a team of engineers, scientists, and medical doctors.

  • Reasoning: The sound, or air pressure variation, coming out from our mouth and nose is intricately tied to changes in anatomy and physiology of the respiratory system. The WHO lists dry cough, difficulty in breathing, and chest pain as key symptoms of COVID-19, visible between 2-14 days after exposure to the virus.
  • Goals: a) To publish a large open-access dataset of respiratory sound samples, (b) design signal processing and machine learning techniques to analyze and detect respiratory infections, and (c) augment expertise from doctors to create effective, easy to use, widely accessible, and cost-effective tools for remote diagnosis of respiratory infections using respiratory sound samples.
  • Note: We do not intend to replace the existing COVID-19 chemical testing methodologies but to supplement them with cost effective, fast and simpler techniques for pre-screening of Covid-19.

This blog will share our journey in this project. We will make posts as and when we find something worth sharing. This will include tutorials, literative review, and updates on the dataset.

Will you like to contribute your sound samples to the database? Click here

Will you like to access the database? click here

Will you like to see an overview video on the project? click here

This site is built with fastpages, an easy to use blogging platform with extra features for Jupyter Notebooks.