Researchers at North Carolina State University have developed a more accurate way to detect coughing using wearable health monitors. This innovation could improve how doctors monitor chronic respiratory conditions like asthma and help track symptoms of infectious diseases such as COVID-19 or the flu.
Traditional cough detection systems rely mostly on audio recordings. However, these systems often misidentify other sounds, such as speech, sneezes, or throat clearing, as coughs. This leads to false positives and limits the usefulness of the technology in real-world environments. To address this, the NC State team created a system that combines audio data with motion data from an accelerometer worn on the chest. The accelerometer captures the physical movement of the chest during a cough, which helps the system distinguish actual coughs from other noises.
The researchers also added a feature called out-of-distribution detection. This allows the system to recognize when it encounters unfamiliar sounds that it was not trained on. Instead of trying to classify these unknown sounds as coughs or non-coughs, the system flags them as uncertain. This helps reduce errors and makes the model more reliable in everyday settings where unexpected noises are common.
In testing, the system achieved over 90 percent accuracy when using standard audio sampling rates. Even when the audio quality was reduced to protect user privacy, the system still performed well. This is important because lower-quality audio makes it harder to identify specific words or conversations, which helps maintain confidentiality while still allowing for effective health monitoring.
The research team believes this technology could be especially useful for remote health monitoring. Patients with asthma or other chronic respiratory conditions could wear the device at home, allowing doctors to track symptoms over time without requiring frequent office visits. It could also help detect early signs of respiratory infections, enabling quicker treatment and reducing the risk of spreading illness.
Article from NC State: Improved Cough-Detection Tech Can Help With Health Monitoring
Abstract in IEEE Journal of Biomedical and Health Informatics: Robust Multimodal Cough Detection with Optimized Out-of-Distribution Detection for Wearables
