Predicting Sleep Quality through Biofeedback: A Machine Learning Approach Using Heart Rate Variability and Skin...

Area of Science:

• Physiology
• Biomedical Engineering
• Machine Learning

Background:

• Sleep quality (SQ) is vital for health, impacting cognition and chronic disease risk.
• Assessing SQ aids in identifying at-risk individuals and developing interventions.
• Wearables and contactless tech can objectively monitor heart rate variability (HRV) and skin temperature for SQ assessment.

Purpose of the Study:

• To develop machine learning models predicting SQ using HRV and skin temperature metrics during wakefulness.
• To evaluate unimodal and multimodal approaches for SQ prediction.
• To establish a non-intrusive method for continuous SQ monitoring.

Main Methods:

• Collected HRV data using a wearable sensor.
• Measured facial skin temperature via infrared thermal imaging.
• Developed and compared classification models, including Support Vector Machine (SVM), using unimodal and multimodal data.

Main Results:

• The Support Vector Machine model utilizing multimodal HRV and skin temperature achieved the highest classification accuracy of 83.4%.
• This multimodal approach demonstrated superior performance compared to unimodal methods and existing state-of-the-art techniques.
• The findings highlight the synergistic effect of combining HRV and skin temperature for comprehensive SQ assessment.

Conclusions:

• Machine learning models integrating HRV and skin temperature can accurately predict sleep quality using non-intrusive measurements during wakefulness.
• This approach offers a promising avenue for employing wearable and contactless technologies in ergonomic applications for continuous SQ monitoring.
• The study advances objective SQ assessment, providing a robust method with significant potential for widespread use.