An ML-driven framework for continuous wearable health surveillance and energy-efficient communication

Area of Science:

• Biomedical Engineering
• Computer Science
• Wireless Communications

Background:

• Existing wireless body area network (WBAN) communication schemes rely on fixed event assumptions, leading to inefficient energy use and scalability issues in continuous health monitoring.
• Redundant transmissions and poor energy utilization hinder the effectiveness of current WBAN systems for real-time health data acquisition.

Purpose of the Study:

• To develop a machine learning-assisted (ML-assisted) framework for WBANs to enhance energy efficiency and communication scalability.
• To enable selective and energy-efficient communication in WBANs through data-driven transmission probability estimation.
• To reduce redundant transmissions and improve channel utilization in continuous health monitoring systems.

Main Methods:

• Developed an ML-assisted Bitmap-Assisted MAC framework utilizing data-driven transmission probability estimation.
• Trained machine learning models in MATLAB using wearable health data to predict the relevance of data transmissions.
• Integrated estimated transmission probabilities into the CC2420 ZigBee energy model for MAC-layer analysis.

Main Results:

• The proposed ML-assisted framework selectively activates transmission-relevant BAN devices, suppressing non-critical traffic.
• Achieved a 35-60% reduction in energy consumption compared to conventional MAC schemes.
• The Gradient Boosting-assisted variant demonstrated superior performance in probability estimation accuracy and overall energy efficiency.

Conclusions:

• The ML-assisted Bitmap-Assisted MAC framework offers a significant improvement in energy efficiency and scalability for WBANs.
• Data-driven transmission probability estimation is effective in optimizing communication and reducing energy waste in continuous health monitoring.
• The proposed approach addresses the limitations of fixed event assumptions in existing WBAN communication schemes.