三流时空步态神经网络用于Kinect身份鉴定  

Three Stream Spatiotemporal Gait Neural Network for Kinect Identity Authentication

作　　者：严云飞 YAN Yunfei(College of Computer and Cyber Security,Fujian Normal University,Fuzhou,China,350117)

机构地区：[1]福建师范大学计算机与网络空间安全学院,福州350117

出　　处：《福建电脑》2025年第1期1-10,共10页Journal of Fujian Computer

摘　　要：步态识别是一种生物特征识别技术。为提高步态识别在走廊场景下的准确率,本文提出一种利用三流图结构数据的混合深度学习模型。该模型基于时间卷积网络和空间卷积网络,以充分提取步态图中最具代表性的时空步态动态特征。通过在30名受试者的正常步态数据上的实验研究,结果表明,该方法的最大准确率达到99.58%,说明该方法能够学习到步态时空的差异特征。Gait recognition is a biometric recognition technology.To improve the accuracy of gait recognition in corridor scenes,this paper proposes a hybrid deep learning model using three stream graph structured data.The model is based on both temporal and spatial convolutional networks to fully extract the most representative spatiotemporal gait dynamic features from gait maps.Through experiments on normal gait data of 30 subjects,the results showed that the maximum accuracy of this method reached 99.58%,indicating that the method can learn spatiotemporal differences in gait features.

关 键 词：时空步态分析 身份鉴定 图神经网络 特征提取 

分 类 号：TP389.1[自动化与计算机技术—计算机系统结构]