By analyzing facial features to perform expression recognition and health monitoring, facial perception plays a pivotal role in noninvasive, real-time disease diagnosis and prevention. Current perception routes are limited by structural complexity and the necessity of a power supply, making timely and accurate monitoring difficult. Herein, a self-powered poly(vinyl alcohol)-gellan gum-glycerol thermogalvanic gel patch enabling facial perception is developed for monitoring emotions and atypical pathological states. Due to the high thermopower of 1.89 mV/K as well as excellent stretchability of 680%, the on-face-conformed thermoelectric gel can operate upon facial thermoelectric variation resulting from different interfacial contact statuses between the gel and face induced by facial muscle activities. With the aid of machine learning, the patch array delivers accurate perception of facial activities of 11 muscles in real time, achieving active expression recognition and health monitoring with the accuracy of 98 and 96%, respectively. This work provides a promising strategy of actively monitoring multisite physiological activities, advancing the development of intelligent wearable bioelectronics for physical or mental monitoring.
Keywords: expression recognition; facial perception; health monitoring; self-powered; thermoelectric gel.