Research · 2025 · Completed
Cough Audio for Health Screening
Cough detection, localization, and tuberculosis risk classification
Machine learning on cough audio: detecting and localizing cough events in continuous recordings, and classifying tuberculosis risk with transformer-based audio models.
Role: Student Researcher · Ubiquitous Computing Lab, UW
- audio
- health
- transformers
Overview
Work at the Ubiquitous Computing Lab at the University of Washington on screening for disease from cough sounds.
- Cough detection and localization: models that find cough events and mark their timing in continuous audio recordings.
- Tuberculosis screening: transformer-based architectures, including the Audio Spectrogram Transformer (AST) and MobileViT, applied to cough-based tuberculosis risk classification.