Intelligent Estimation of Exercise Induced Energy Expenditure Including Excess Post-Exercise Oxygen Consumption...

Area of Science:

• Exercise Physiology
• Biomedical Engineering
• Machine Learning Applications

Background:

• Calorimetry systems for measuring human energy expenditure (EE) during exercise are limited in availability.
• Wearable sensors offer a promising alternative for accessible EE estimation.
• Machine learning can integrate various physiological data for improved EE prediction.

Purpose of the Study:

• To develop and validate a novel method for estimating human energy expenditure (EE) during aerobic exercise using wearable sensors and machine learning.
• To incorporate both exercise energy consumption and excess post-exercise oxygen consumption (EPOC) into the estimation model.
• To assess the accuracy of the proposed method against established ground truth measurements.

Main Methods:

• Thirty-two healthy adults performed 20-minute aerobic exercise sessions at low (40% VO2 max) and high (70% VO2 max) intensities.
• Physical characteristics, exercise intensity, and heart rate data were collected from exercise onset until resting metabolic rate recovery.
• Machine learning algorithms were employed to build EE estimation models using the collected data.

Main Results:

• The developed EE estimation method demonstrated a high correlation coefficient of up to 0.976 with ground truth values.
• The root mean square error (RMSE) for the estimation was 0.624 kcal/min, indicating precise predictions.
• The models effectively integrated exercise intensity and heart rate data for accurate EE calculation.

Conclusions:

• A highly accurate wearable sensor-based method for estimating human energy expenditure during exercise was successfully developed.
• The method's high correlation with ground truth validates its potential for practical applications.
• This technology can significantly benefit fitness tracking, healthcare monitoring, and sports performance analysis.