基于运动特征的骨骼行为识别方法  

Human skeleton activity recognition based on motion feature

作　　者：孙浩[1] 何宏 汪焰兵 朱子豪 SUN Hao;HE Hong;WANG Yan-bing;ZHU Zi-hao(School of Health Science and Engineering,University of Shanghai for Science and Technology,Shanghai 200093,China)

机构地区：[1]上海理工大学健康科学与工程学院,上海200093

出　　处：《计算机工程与设计》2024年第6期1836-1842,共7页Computer Engineering and Design

基　　金：国家科技部基金项目(G2021013008、2020YFC2005802);上海市科学技术委员会基金项目(18070503000);上海理工大学医工交叉重点基金项目(1020308405、1022308502)。

摘　　要：针对现有的骨骼行为识别方法对人体行为的运动信息利用不足的问题,提出一种基于运动特征的时空注意力图卷积(STA-GCN)行为识别模型。对动作捕捉设备采集到的关节点运动轨迹和速度信息进行建模,在时间和空间构建注意力权重矩阵,结合图卷积网络进行特征提取,能够关注到具有判别力的关节点和时间帧。通过在自建动作捕捉数据集和NTU-RGB+D数据集的CS和CV标准上进行实验,其结果表明,该模型增强了对人体骨骼行为信息的理解能力,验证了模型对行为识别的有效性。Aiming at the problem that the existing skeleton-based activity recognition algorithms can not make full advantage of motion information of human activity,a spatial-temporal attention graph convolutional neural network(STA-GCN)based on motion feature was proposed to model the trajectory and velocity collected by motion capture system.The spatial and temporal weight matrix was constructed and it was combined with graph convolutional network for feature extraction to focus the discriminative joints and frames.Experimental results on dataset established using activity information collected by motion capture system and NTU-RGB+D dataset’s both CS and CV evaluation criterion show that the model enhances the ability to understand the information of human skeleton activity,and its effectiveness is verified for activity recognition.

关 键 词：行为识别 深度学习 动作捕捉 骨骼信息 特征提取 图卷积 时空注意力 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]