A machine learning-integrated multi-channel flexible optical fiber wearable system for high-accuracy human motion monitoring in dynamic environments

Traditional single-sensor systems face inherent limitations in achieving high-accuracy motion monitoring with low latency, particularly under complex dynamic conditions. To overcome these constraints, this work introduces a multi-channel flexible optical fiber wearable system integrated with machine learning for human motion analysis. Leveraging flexible optical fiber sensor arrays and a 1D convolutional neural network (1D-CNN) architecture, our system automatically extracts discriminative motion features from strain and pressure signals, significantly enhancing detection accuracy. Two application-oriented prototypes were developed: a smart glove for gesture recognition and a smart carpet for full-body motion monitoring. Experimental validation demonstrates exceptional performance, with 98.12 % classification accuracy for eight distinct gestures and 96.82 % recognition accuracy for six fundamental body movements. This innovation establishes a robust platform for quantitative motion assessment, offering promising potential in rehabilitation medicine (motor function recovery tracking), sports biomechanics (posture optimization analysis), and immersive human-computer interaction systems (virtual reality deformation sensing). The proposed methodology advances wearable sensing technology by synergizing optical signal transduction with adaptive deep learning frameworks.