Core-shell hydrogel fiber sensor-integrated smart glove enabling machine learning-assisted gesture recognition

Gesture recognition is fundamental to human-computer interaction (HCI), yet most existing approaches depend on multi-modal sensing systems that impose substantial hardware complexity and computational overhead. Here, we present a single-modal strain sensing strategy for sign language recognition, enabled by a core–shell structured carbon nanotube (CNT)/carbomer hydrogel fiber sensor integrated into a wearable smart glove. Compared with a conventional CNT/poly (vinyl alcohol) hydrogel sensor, the core–shell architecture affords markedly enhanced sensing stability and fatigue resistance. Within this structure, CNTs function simultaneously as conductive fillers, nanoscale mechanical reinforcements, and stress-transfer pathways, stabilizing the hydrogel network and augmenting strain transduction efficiency through modulus mismatch-induced microstructural rearrangement. Sensor placement was further optimized in accordance with finger biomechanics, enabling single-modality strain signals to faithfully capture coordinated multi-joint motion. When coupled with lightweight machine learning classifiers, the resulting smart glove achieves sign language recognition with an accuracy of 98.75%. These findings demonstrate that structural design and biomechanically informed layout optimization can improve the information density of single-modal strain signals, enabling reliable gesture recognition with reduced sensing coverage and a simplified wearable architecture. The proposed framework provides a scalable and wearable solution for flexible electronics and intelligent HCI applications.