SweaTag: Fine-Grained Sweat Amount Sensing with COTS RFID Tags

Accurate sweat amount sensing can provide key insights into sports scenarios (e.g., exercise load monitoring) and medical scenarios (e.g., hyperhidrosis monitoring). However, current research on sweat either focuses on sensing its composition or requires sophisticated sensors to measure sweat amount. To fill this research gap, in this paper, we propose a novel RFID-based solution, Sweatag,to perform sweat amount sensing with COTS RFID tags. When the sweat amount varies, the equivalent impedance formed between the sweat and the tag antenna changes, affecting the corresponding signal phase. Capturing this phase change is challenging under the interference of translation-type and rotation-type moving effect introduced by human motion. To remove the translation-type interference, we propose a tag-pair method by taking the phase difference between two adjacent tags. To remove the rotation-type interference, we propose a multi-frequency differential method based on the observation that the moving effect has an identical impact on two close-by frequencies. To provide multiple frequencies simultaneously, we propose to leverage Orthogonal Frequency Division Multiplexing (OFDM) and provide a whole process to extract the sweat amount fingerprint from the physical layer signals. We use Partial Least Squares Regression (PLSR) to reduce the dimensionality of the fingerprint and estimate the sweat amount. Experiment results show that Sweatagachieves 93% average sensing accuracy for detecting five sweat amount levels.