学术文献库

Staying hydrated and drinking fluids is extremely crucial to stay healthy and maintaining even basic bodily functions. Studies have shown that dehydration leads to loss of productivity, cognitive impairment and mood in both men and women. However, there are no such an existing tool that can monitor dehydration continuously and provide alert to users before it affects on their health. In this paper, we propose to utilize wearable Electrodermal Activity (EDA) sensors in conjunction with signal processing and machine learning techniques to develop first time ever a dehydration self-monitoring tool, \emph{Monitoring My Dehydration} (MMD), that can instantly detect the hydration level of human skin. Moreover, we develop an Android application over Bluetooth to connect with wearable EDA sensor integrated wristband to track hydration levels of the users real-time and instantly alert to the users when the hydration level goes beyond the danger level. To validate our developed tool's performance, we recruit 5 users, carefully designed the water intake routines to annotate the dehydration ground truth and trained state-of-art machine learning models to predict instant hydration level i.e., well-hydrated, hydrated, dehydrated and very dehydrated. Our system provides an accuracy of 84.5% in estimating dehydration level with an sensitivity of 87.5% and a specificity of 90.3% which provides us confidence of moving forward with our method for larger longitudinal study.