计算机科学 ›› 2024, Vol. 51 ›› Issue (4): 236-242.doi: 10.11896/jsjkx.221200120
闫文杰, 尹艺颖
YAN Wenjie, YIN Yiying
摘要: 图卷积神经网络(Graph Convolutional Neural network,GCN)在基于3D骨架的人体行为识别领域取得了良好效果。然而,现有的大多数GCN方法对行为动作图的构建都是基于人体物理结构的手动设置,训练阶段各个图节点只能根据手动设置建立联系,无法感知动作行为过程中骨骼节点之间产生的新联系,导致图拓扑结构不合理和不灵活。移位图卷积网络通过改变图网络结构使得感受野更加灵活,并且在全局移位角度取得了良好效果。因此,提出了一种基于自适应移位图卷积神经网络(Adaptive Shift Graph Convolutional Neural network,AS-GCN)的人体行为识别算法来弥补前述GCN方法的不足。AS-GCN借鉴了移位图卷积网络的思想,提出用每个人体动作的本身特点来指导图神经网络进行移位操作,以尽可能准确地选定需要扩大感受野的节点。在基于骨架的通用动作识别数据集NTU-RGBD上,所提算法在骨骼有无物理关系约束的前提条件下均进行了实验验证。与现有的先进算法相比,AS-GCN算法的动作识别准确率在有骨骼物理约束的条件下的CV和CS角度上平均提高了12%和4.84%;在无骨骼物理约束的条件下的CV和CS角度上平均提高了20%和14.49%。
中图分类号:
[1]KONG W,LIU Y,LI H,et al.A survey of action recognitionmethods based on graph convolutional network [J].Control and Decision,2021,36(7):1537-1546. [2]LIANG X,LI W X,ZHANG H N.Review of research on human action recognition methods [J].Application Research of Computers,2022,39(3):651-660. [3]ZHAO X H,YE S,LI X.Multi-algorithm Fusion Behavior Classification Method for Body Bone Information Reconstruction [J].Computer Science,2022,49(6):269-275. [4]MIAO G Q,XIN W T,LIU R Y,et al.Graph ConvolutionalSkeleton-based Action Recognition Method for Intelligent Behavior Analysis [J].Computer Science,2022,49(2):156-161. [5]ZHANG P,XUE J,LAN C,et al.EleAtt-RNN:Adding atten-tiveness to neurons in recurrent neural networks[J].IEEE Transactions on Image Processing,2019,29:1061-1073. [6]LI M,SUN Q.3D Skeletal Human Action Recognition Using a CNN Fusion Model [J].Mathematical Problems in Engineering,2021(18):6650632.1-6650632.11. [7]LI Y Z,YUAN J Z,LIU H Z.Human skeleton-based action re-cognition algorithm based on spatiotemporal attention graph convolutional network model [J].Journal of Computer Applications,2021,44(7):1915-1921. [8]YAN S,XIONG Y,LIN D.Spatial temporal graph convolutional networks for skeleton-based action recognition[C]//Thirty-se-cond AAAI Conference on Artificial Intelligence.Palo Alto,Cali-fornia USA:AAAI Press,2018:7444-7452. [9]CHENG K,ZHANG Y,HE X,et al.Skeleton-based action re-cognition with shift graph convolutional network[C]//Procee-dings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.San Francisco,CA,USA:IEEE,2020:183-192. [10]SHI L,ZHANG Y,CHENG J,et al.Two-stream adaptive graph convolutional networks forskeleton-based action recognition[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.San Francisco,CA,USA:IEEE,2019:12026-12035. [11]SHI L,ZHANG Y,CHENG J,et al.Skeleton-based action re-cognition with multi-stream adaptive graph convolutional networks [J].IEEE Transactions on Image Processing,2020,29:9532-9545. [12]SHAHROUDY A,LIU J,NG T T,et al.Ntu rgb+ d:A large scale dataset for 3d human activity analysis[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.San Francisco,CA,USA:IEEE,2016:1010-1019. [13]LI B,LI X,ZHANG Z,et al.Spatio-temporal graph routing for skeleton-based action recognition[C]//Proceedings of the AAAI Conference on Artificial Intelligence.Palo Alto,California USA:AAAI Press,2019,33(01):8561-8568. [14]ZHANG P,LAN C,ZENG W,et al.Semantics-guided neuralnetworks for efficient skeleton-based human action recognition[C]//proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.San Francisco,CA,USA:IEEE,2020:1112-1121. [15]PENG W,HONG X,CHEN H,et al.Learning graph convolutional network for skeleton-based human action recognition by neural searching[C]//Proceedings of the AAAI Conference on Artificial Intelligence.Palo Alto,California USA:AAAI Press,2020,34(3):2669-2676. [16]CHENG K,ZHANG Y,CAO C,et al.Decoupling gcn withdropgraph module for skeleton-based actionrecognition[C]//European Conference on Computer Vision.Cham:Springer,2020:536-553. |
|