Non-invasive gesture recognition using PDMS-encapsulated polymer optical fiber Bragg gratings

This paper introduces a non-invasive and high-sensitivity method for gesture recognition based on polymer optical fiber Bragg grating (POFBG) sensors embedded in polydimethylsiloxane (PDMS). Experimental results show that PDMS-encapsulated POFBGs exhibit bending sensitivities for finger and wrist motions that are 8.3 times that of conventional silica optical fiber Bragg gratings, while maintaining excellent repeatability. For gesture recognition, four POFBG sensors are strategically attached to specific forearm muscles to monitor muscle deformations produced by various gestures. The response characteristics and sensing performance of the sensors are comprehensively validated. A convolutional neural network–long short-term memory–attention model is employed to train and predict 16 gestures, including five static wrist gestures, eight static finger gestures, and three dynamic gestures. The model effectively discriminates subtle wavelength shifts caused by fine finger movements, achieving an overall recognition accuracy of 95.85% across different subjects. The proposed approach enables integrated and precise recognition of both static and dynamic gestures using a minimal number of sensors, achieving an optimal balance between recognition accuracy, coverage, and device portability. This work provides a high-performance and cost-effective sensing solution for advanced human–computer interaction applications.