Activity-Aware Wearable System for Power-Efficient Prediction of Physiological Responses

Area of Science:

• Biomedical Engineering
• Wearable Technology
• Health Informatics

Background:

• Wearable health monitoring offers remote assessment and personalized care.
• Improving energy efficiency and battery life is crucial for wearable device adoption.
• Self-powered wearable devices are essential for continuous health monitoring.

Purpose of the Study:

• To propose a context-aware dynamic sensor selection method for power-optimized physiological prediction.
• To reduce energy consumption in multi-modal wearable data streams.
• To enhance prediction accuracy through intelligent sensor management.

Main Methods:

• Clustering physiological data by physical activity using accelerometer data.
• Applying group lasso models to activity clusters for feature selection.
• Implementing duty cycling for selected sensors to reduce power usage.

Main Results:

• Activity state-based contextual information improved prediction accuracy and decreased power usage.
• A reduced feature set enhanced accuracy and decreased energy burden in regression models.
• Demonstrated significant power usage reduction with a custom multi-modal wearable prototype.

Conclusions:

• Context-aware dynamic sensor selection is effective for power optimization in wearable health monitoring.
• This approach enhances the feasibility of long-term, continuous physiological data collection.
• The method holds potential for improving patient outcomes through efficient wearable technology.