Machine Learning Approaches for Exercise Exertion Level Classification Using Data from Wearable Physiologic Monitors.

Area of Science:

• Exercise Physiology
• Biomedical Engineering
• Data Science

Background:

• Monitoring aerobic exercise exertion is crucial for optimizing training and preventing overexertion.
• Heart rate variability (HRV) offers a non-invasive physiological marker of autonomic nervous system activity during exercise.
• Real-time feedback on exertion levels can enhance exercise adherence and performance.

Purpose of the Study:

• To develop a predictive model for real-time aerobic exercise exertion levels.
• To identify key heart rate variability features for classifying exertion states.
• To evaluate the performance of machine learning models in predicting exertion.

Main Methods:

• ECG signals were recorded during 16-minute cycling exercises.
• Perceived ratings of exertion (RPE) were collected to label exercise minutes as 'high' or 'low' exertion.
• Time and frequency domain HRV features were extracted and ranked using the mRMR algorithm.
• A support vector machine classifier was trained and evaluated.

Main Results:

• The minimum redundancy maximum relevance (mRMR) algorithm identified the top ten predictive HRV features.
• The support vector machine model achieved the highest classification accuracy.
• An F1 score of 82% was obtained for predicting high versus low exertion levels.

Conclusions:

• HRV characteristics derived from ECG signals can effectively predict aerobic exercise exertion levels in real-time.
• Machine learning models, particularly support vector machines, are suitable for developing such predictive systems.
• This approach holds potential for personalized exercise guidance and performance monitoring.