IMPROVING RELIABILITY IN WEARABLE HEALTH MONITORING VIA MACHINE LEARNING

Dina Hussein

Wearable Internet of Things (IoT) devices are transforming healthcare by enabling continuous and personalized monitoring of physiological and behavioral signals. These devices can detect anomalies, support early interventions, and empower patients to manage chronic conditions in daily life. However, the widespread adoption of wearable health technologies is limited by three fundamental challenges: (i) constrained energy resources that shorten device lifetime and necessitate frequent recharging, (ii) unreliable sensing conditions leading to missing or disturbed data that degrade application accuracy, and (iii) the need for sensor- and energy-aware algorithms that can operate robustly under dynamic user and environmental conditions. Addressing these challenges is critical to realizing the promise of intelligent and reliable health monitoring systems. This dissertation presents a suite of novel algorithms for energy-efficient, robust, and sensor-aware wearable IoT systems. Contributions span three complementary research pillars that collectively enable reliable and sustainable operation of health monitoring applications.First, we develop generative and statistical models to recover missing or corrupted sensor data, ensuring the reliability of health monitoring even in the face of degraded sensing conditions. These approaches prioritize application-specific trade-offs between computational cost and reconstruction fidelity. Second, we design adaptive energy management strategies that empower wearable devices to operate sustainably despite unpredictable energy availability. By integrating lightweight decision-making and energy harvest forecasting, these techniques enable continuous system operation without compromising performance. Third, we introduce sensor-aware and energy-efficient techniques that adjust sensing and computation tasks dynamically based on energy levels. These contributions advance the design of intelligent, self-sustaining, and reliable wearable IoT devices for health monitoring. By systematically addressing core challenges in data reliability, energy sustainability, and sensor-aware computation through the proposed approaches, this work lays the foundation for next-generation systems that operate continuously in real-world conditions while enabling equitable access to personalized healthcare monitoring.

IMPROVING RELIABILITY IN WEARABLE HEALTH MONITORING VIA MACHINE LEARNING

Files and links (1)

Abstract

Metrics

Details