Acoustic Detection of COVID-19

Especially interesting as an early-onset detection and screening.

Research papers

June 20, 2020. Exploring Automatic Diagnosis of COVID-19 from Crowdsourced Respiratory Sound Data. Chloë Brown et.al, Cambride University. Crowdsourced dataset of respiratory sounds Used low-level acoustical features , and MFCCs. To date 10,000 samples from about 7000 unique users. Tested cough classification into COVID-19 positive/negative. AUC ROC over 70%. Also tested using breathing sounds.

Data

covid-19-sounds.org. Crowdsourced data collection effort by Cambridge researchers. Using an Android and iOS app. Collecting both coughing and breathing, from healthy and non-healthy users. No information on availability as of June 2020.

In the Media

May 12, 2020. Researchers at Pitt and clinicians at the University of Pittsburgh Medical Center Children’s Hospital https://www.engineering.pitt.edu/News/2020/Smartphone-Diagnostics
April 15, 2020. Researchers from Indian Institute of Science (IISc), Bangalore, developing an acoustic diagnostic tool for COVID-19 based on cough and speech sounds called "Coswara". No data collection or deployment strategy mentioned. Link