Machine-learning-assisted wearable PVA/Acrylic fluorescent layer-based triboelectric sensor for motion, gait and individual recognition

Abstract

Flexible sensors with precise perception ability play an important role in environmental sensing. The organic fusion between the sensitive structure and sensitive material presents some challenges for the overall design and integration of the sensor, including designing multiple sensitive branch structures, simplifying system design, and sensing multiple stress modes. Here, based on a same-plane interdigital electrode design, we developed a triboelectric sensor and its array (TSA) that mimics the synaptic structure of a neural cell. The interdigital electrode structure can effectively simulate the synaptic structure of neural cells and achieve effective perception of multi-directional stress. The enhancement of the triboelectric effect is achieved through layer-by-layer self-assembly technology between PVA film and acrylic fluorescent layer. The interface effect between PVA film and acrylic fluorescent layer can effectively enhance the electrostatic effect. Further analysis with deep learning CNN can extract higher-level motion features. A single sensor can be used to identify which part of the body is producing the movement (98.89%). A TSA with different numbers of sensors can achieve high-precision recognition of gesture numbers (99.25%), foot shape, gait (99.23%), and individuals (99.75%). The sensor can be flexibly manufactured and integrated according to specific application scenarios. The long-term goal is to effectively help limb-injured patients with rehabilitation therapy and build intelligent home environment.