A Non-Invasive, Machine Learning Assisted Skin-Hydration Microwave Sensor

This paper presents a non-invasive and electrodeless skin-hydration sensor based on a microwave resonator, assisted by a machine learning (ML) classification algorithm. A miniaturized complementary-split ring resonator (CSRR) operating in the 2÷3 GHz band is designed to sense and detect hydration changes in the human skin utilizing the near-field interaction with the resonator. The resonator is first simulated using electromagnetic (EM) simulations and loaded with skin in different hydration conditions. Then, the resonator is fabricated and tested in three different body regions: thenar eminence, proximal wrist ceases, and cheek. Spectra are collected during an extensive experimental campaign in which repeated measurements are taken over three days on the three regions of the body. The Soft Independent Modelling of Class Analogy (SIMCA) method is then used to interpret and classify the spectral data acquired from the resonator into two distinct hydration classes. The results of this study highlight the effectiveness of SIMCA as a multivariate analysis technique for processing and categorizing spectral data obtained from CSRR for monitoring body hydration. This approach provides a cost-effective solution and demonstrates high efficiency in accurately distinguishing between hydration states, which holds great promise for practical applications in hydration monitoring.