Coswara: speech and sound based diagnostics

Team Lead: Sriram Ganapathy

COVID-19 is an infectious disease caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2).  One of the most common symptoms of COVID-19 is persistent dry cough, with others including presence of  respiratory sputum (phlegm) and shortness of breath.

The only approved form of diagnosis for viral infection is reverse transcription polymerase chain reaction (RT-PCR) of infected secretions. A real time RT-PCR test is the most commonly deployed test currently for COVID-19 with testing results available after several hours. Generally, this test is carried out using a nasopharyngeal swab or a throat swab. The major limitations of the testing which hinder the widespread deployment of this procedure include violation of social distancing which increases the chance of infection to the subject, exposure of the viral infection to healthcare workers, expenses involved in chemical testing and the time required for results with an increasing population requiring the test. As the pandemic is growing in numbers, the development of simplistic, cost-effective and fast testing for infection has become a crucial component in healthcare, policy making and economic revival of several countries.

This project, named ”Coswara” (, attempts to provide a simple tool for diagnostics of COVID-19 based on respiratory, cough and speech sounds. As the major symptoms of the disease include respiratory problems, the proposed project aims to detect and quantify the biomarkers of the disease in the acoustics of these sounds. The project requires participants to perform a recording of breathing sounds, cough sounds, sustained phonation of vowel sounds and a counting exercise. The entire response requires about 5 minutes of recording time. Along with these recordings, the tool also records the patient’s health status (without any personally identifiable information) as well as age, gender and location. The audio dataset collected will be released for researchers across the world to develop a potential diagnostic tool using signal processing and machine learning methods.

The project is in the data collection stage and will go through an experimental validation before obtaining full approval as a potential diagnostic tool. Given the highly simplistic and cost-effective nature of the tool, we hypothesize that, even a partial success for the tool would enable a massive deployment as a first line diagnostic tool for the pandemic. The project is not aimed to replace the chemical testing or the imaging methods but to merely supplement those with a cost-effective, fast and simple technique.  

The webpage for data collection is here:


  • Anand Mohan (IISc MTech alumnus)
  • Neeraj Sharma (Postdoc at CMU and IISc PhD alumnus from ECE)
  • Srikanth Raj (Postdoc and PhD at IISc)
  • Shreyas Ramoji (PhD student, IISc)
  • Sriram Ganapathy (faculty, EE, IISc)
  • Prasanta Ghosh (faculty, EE, IISc)
  • Shahbhaz Sultan (Cogknit Semantics)
  • Anuroop Iyengar (CEO, Cogknit Semantics)

Scroll Up