HydroTrack: Spectroscopic Analysis Prototype Enabling Real-Time Hydration Monitoring in Wearables
Nazim A. Belabbaci, Mohammad Arif Ul Alam
TL;DR
Hydration monitoring in wearables is challenging due to the need for real-time, non-invasive assessment. The authors introduce HydroTrack, a smartwatch-integrated 18-channel spectrophotometer with on-device edge processing and Euler-based signal magnification to detect hydration states, validated on six participants with up to 95% accuracy. They demonstrate two use cases—electrolyte-solution spectroscopy and skin hydration during exercise—and show edge deployment of a Random Forest classifier on a DSTike watch, with XGBoost achieving higher performance in several analyses. This work advances practical, real-time optical hydration sensing in wearables and lays groundwork for personalized, color-aware hydration monitoring on resource-constrained devices.
Abstract
In the rapidly growing field of wearable technology, optical devices are emerging as a significant innovation, offering non-invasive methods for analyzing skin and underlying tissue properties. Despite their promise, progress has been slowed by a lack of specialized prototypes and advanced analysis techniques. Addressing this gap, our study introduces, HydroTrack, an 18-channel spectroscopy sensor, ingeniously embedded in a smart-watch. Accompanying this hardware, we present signal processing and data analysis techniques implemented at the edge, designed to maximize the utility of our system in comprehensive health tracking. A pivotal application of our device is the real-time assessment of hydration levels in physically active individuals. We validated our prototype and analytical approach through experiments on six participants, focusing on hydration dynamics during physical exercises. Our findings reveal an accuracy of avg. 95% in determining hydration states.
