用于篮球持球姿势识别与训练的石墨烯/凯夫拉柔性智能传感手套设计与研究

    Smart Gloves Based on Graphene/Kevlar Composite Flexible Sensors for Basketball Grip Posture Recognition and Athletic Training

    • 摘要: 本文针对柔性可穿戴传感器在运动监测和人机交互中对高灵敏度、快速响应及智能识别的需求,研究一种基于凯夫拉针织布的双模态柔性织物传感器,并探索其在智慧体育中的应用。本文以凯夫拉针织布为柔性基底,通过浸涂聚酰亚胺(Polyimide, PI)溶液并结合激光直写技术,诱导聚酰亚胺原位石墨化,在织物表面构建三维多孔激光诱导石墨烯(Laser-induced Graphene, LIG)导电网络。采用扫描电子显微镜(Scanning Electron Microscope, SEM)、热重分析(Thermogravimetric Analysis, TGA)和拉曼光谱对导电层进行表征,并通过拉伸和压力测试评价传感性能。在此基础上,本文将传感器阵列集成于智能手套,采集不同篮球持球姿势下的多通道电阻信号,并采用一维卷积神经网络(1D Convolutional Neural Network, 1D-CNN)进行姿势识别。本文制备的传感器具有良好的拉伸应变和压力双模态感知能力。拉伸传感模式下,应变灵敏系数(Gauge Factor, GF)最高达到48.5,响应时间和恢复时间分别为0.25 s和0.22 s;压力传感模式下,压力灵敏度(Sensitivity, S)最高达到1.22 kPa−1,响应时间和恢复时间分别为0.03 s和0.05 s。两种传感模式在持续数千秒的长期循环测试过程中均表现出稳定的响应。基于智能手套采集的四种典型篮球持球姿势信号,一维卷积神经网络实现了94%的姿势识别准确率。本文所制备的柔性织物传感器可实现应变和压力双模态感知,结合智能手套与深度学习可实现篮球持球姿势识别,在智慧体育和智能可穿戴领域具有应用潜力。

       

      Abstract: This study addresses the demand for high sensitivity, rapid response, and intelligent recognition in flexible wearable sensors for motion monitoring and human-machine interaction. A dual-modal flexible textile sensor based on knitted Kevlar fabric was developed, and its application in smart sports was investigated. Knitted Kevlar fabric was used as the flexible substrate, and a polyimide (PI) solution was applied by dip coating, followed by laser direct writing to induce in situ graphitization of polyimide and construct a three-dimensional porous laser-induced graphene (LIG) conductive network on the fabric surface. Scanning electron microscopy (SEM) , thermogravimetric analysis (TGA) , and Raman spectroscopy were employed to characterize the conductive layer, while tensile and pressure tests were conducted to evaluate the sensing performance. Subsequently, a sensor array was integrated into a smart glove to collect multichannel resistance signals under different basketball-holding postures, and a one-dimensional convolutional neural network (1D-CNN) was employed for posture recognition. The fabricated sensor exhibited good dual-modal sensing capability for tensile strain and pressure. In the tensile sensing mode, the gauge factor (GF) reached a maximum of 48.5, with a response time and recovery time of 0.25 s and 0.22 s, respectively. In the pressure sensing mode, the sensitivity (S) reached a maximum of 1.22 kPa−1, with a response time and recovery time of 0.03 s and 0.05 s, respectively. Both sensing modes maintained stable responses during long-term cyclic testing lasting thousands of seconds. Based on the signals collected by the smart glove for four typical basketball-holding postures, the one-dimensional convolutional neural network achieved a posture recognition accuracy of 94%. The fabricated flexible textile sensor enables dual-modal strain and pressure sensing, and its integration with a smart glove and deep learning enables recognition of basketball-holding postures, demonstrating its potential for applications in smart sports and intelligent wearable devices.

       

    /

    返回文章
    返回