Sensor data quality processing for vital signs with opportunistic ambient sensing.

Area of Science:

• Biomedical Engineering
• Sensor Technology
• Physiological Monitoring

Background:

• Opportunistic ambient sensing enables unobtrusive collection of physiological signals for vital signs monitoring.
• Extracting reliable vital signs from intermittently contacted sensors presents a significant challenge.

Purpose of the Study:

• To develop and evaluate a quality processing system for extracting heart rate from ballistocardiogram signals.
• To assess the effectiveness of various classifiers in identifying informative signals for vital signs monitoring.

Main Methods:

• Utilized a micro-bending fiber optic sensor pressure mat to collect ballistocardiogram signals.
• Employed visual inspection for data labeling and five classifiers (Random Forest, SVM, etc.) for signal classification.
• Processed informative signals to estimate interbeat intervals and compared with reference ECG.

Main Results:

• The quality processing system successfully identified 57.37% of the dataset as informative signals.
• The Random Forest classifier achieved a high accuracy of 98.99% for informative signal classification.
• The mean absolute error in heart rate estimation was reduced from 13.2 to 8.47 compared to reference ECG.

Conclusions:

• The proposed quality processing system demonstrates robustness for opportunistic ambient sensing.
• This approach enhances the reliability of vital signs monitoring using unobtrusive sensor technology.