Hand gesture recognition based on an optical fiber specklegram sensor

Gesture recognition offers considerable promise for enhancing the naturalness, intuitiveness, and efficiency of human-computer interaction. In this paper, a novel hand gesture recognition system based on a single-mode, multimode, multicore optical fiber sensing structure is presented. By leveraging deep learning to analyze fiber specklegrams and training a simplified six-layer ResNet network, accurate and efficient hand gesture recognition was realized with minimal hardware requirements and high stability. Its compact design further supports integration into wearable devices, highlighting its significance for practical applications.