A Novel Multi-Centroid Template Matching Algorithm and Its Application to Cough Detection.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Wearable Technology

Background:

• Coughing is a key symptom in respiratory diseases like COPD and asthma.
• Existing cough detection methods primarily rely on audio analysis.
• Inertial Measurement Units (IMUs) offer a novel, untapped avenue for cough detection through motion analysis.

Purpose of the Study:

• To develop and validate a novel self-tuning multi-centroid template-matching algorithm for cough detection using IMU data.
• To explore the feasibility of using head motion captured by IMUs for identifying cough events.
• To optimize cough detection algorithms for improved accuracy and reduced computational time.

Main Methods:

• A novel self-tuning multi-centroid template-matching algorithm was developed, automatically adjusting cluster numbers.
• Experiments utilized synthetic datasets and a real-world dataset from earbud-based IMU sensors.
• The algorithm was evaluated for its ability to balance accuracy and inference time compared to existing methods.

Main Results:

• The proposed self-tuning multi-centroid algorithm demonstrated superior performance in cough detection.
• Effective cough detection was achieved using a single accelerometer sensor on an earbud platform.
• The algorithm successfully balanced accuracy and inference time, outperforming traditional methods like KNN with DTW.

Conclusions:

• Cough detection using IMU sensors, particularly from ear-worn devices, is a viable and promising approach.
• The novel self-tuning multi-centroid algorithm provides an efficient and accurate solution for IMU-based cough detection.
• This technology has significant potential for remote patient monitoring and management of pulmonary diseases.