Exploring Automatic Diagnosis of COVID-19 from Crowdsourced Respiratory Sound

Explore the potential of using crowdsourced respiratory sound data for automatic diagnosis of COVID-19 in this 21-minute conference talk from KDD 2020. Delve into the process of collecting audio signals, analyzing demographics, and correlating symptoms with COVID-19. Learn about the experimental setup, including global and frame-level features, and discover initial findings on COVID-19 positive detection. Gain insights into data augmentation techniques and their impact on the results. Understand the challenges and opportunities in leveraging audio data for disease diagnosis, presented by researchers from the Association for Computing Machinery (ACM).

Introduction
Audio Signals for COVID-19
Crowdsourcing Data: Collection
Demographics: Country & Age
Symptom Correlation: Broad picture
Symptom Correlation: COVID-19 specific
COVID-19 Positive Detection
Tasks & Data
Global Features
Frame-level Features
Feature Types
Experimental Setup
Initial Findings...
with Data Augmentation
Conclusion