Hand-Size-Adaptable FBG Wearable Glove for Cross-User Gesture Recognition

Gesture recognition is fundamental to intuitive human-machine interaction. However, conventional glove-based systems commonly suffer from performance degradation in multi-user scenarios due to variations in hand size and the resulting sensor misalignment. To address these limitations, this paper presents a hand-size-adaptive wearable glove with five multi-node fiber Bragg grating (FBG) sensors and an associated adaptive gesture recognition method. The glove integrates a distributed sensing layout with an adjustable skeletal glove structure to improve joint coverage across different hand sizes, and a PDMS-TPU composite encapsulation is used to protect the fibers while maintaining sensitivity and flexibility. An incremental update strategy based on an online linear support vector machine (SVM) trained with stochastic gradient descent (SGD) is employed for cross-user adaptation, updating the classifier using a small calibration set from each new user. Experiments on a multi-user gesture dataset show that fine-tuning with only 20% of a new user’s data yields average recognition accuracies of 99.56% for static gestures and 98.83% for dynamic gestures. The combination of adaptive hardware integration and incremental classifier updating supports low-effort, high-accuracy gesture interaction in shared-use wearable applications.