提取一种融合注意力和多尺度时空图网络的人体行为识别算法,在时空图网络卷积层融入通道-空间级联注意力机制以及在时间图卷积中增加多尺度卷积,利用改进的算法通过嵌入式平台在NTU RGB+D数据集的两个评估基准X-Sub和X-View上的准确率达到了89.1%和92.5%。实验结果表明,该方法具有可靠的精度,可以应用于嵌入式平台完成人体行为识别任务。
Abstract
In the paper,a human behavior recognition algorithm is proposed,that combines attention and multi-scale spatiotemporal map network.The channel space cascade attention mechanism is integrated into the space-time map network convolution layer and the multi-scale convolution is added to the time map convolution.The accuracy of the improved algorithm on the two evaluation benchmarks X-Sub and X-View of NTU RGB+D dataset through the embedded platform has reached 89.1% and 92.5%.The experiment results show that the method has reliable accuracy.It can be applied to the embedded platform to complete the task of human behavior recognition.
关键词
ST-GCN /
注意力机制 /
多尺度时间卷积 /
嵌入式平台 /
人体行为识别
Key words
ST-GCN /
attention mechanism /
multi-scale time convolution /
embedded equipment /
human behavior recognition
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] Jiang W,Liu Z,Ying W,et al.Mining actionlet ensemble for action recognition with depth cameras[C]//Proceedings of the 2012 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).IEEE,2012.
[2] Y Du,W Wang,L Wang.Hierarchical recurrent neural net-work for skeleton based action recognition[C]//in Proc.IEEE Conf. Comput. Vis. Pattern Recognit.(CVPR),2015.
[3] V Veeriah,N Zhuang,G J Qi.Differential recurrent neural networks for action recognition[C]//in Proc. IEEE Int. Conf. Comput. Vis.(ICCV),2015.
[4] SI c,CHEN w,WANG w,et al.An attention enhancedgraph convolutional ISTM network for skeleton-based action recognition[C]//IEEE/CVF Conference on Computer Vision and Pattern Recognition,2019.
[5] YAN Sijie,XIONG Yuanjun,LIN Dahua.Spatial temporal graph convolutional networks for skeleton-based actionrecognition[J].2018.
[6] Li M,Chen S,Chen X,et al.Actional-Structural Graph Convolutional Networks for Skeleton-based Action Recognition[J].IEEE,2019.
[7] Shi L,Zhang Y,Cheng J,et al.Two-Stream Adaptive Graph Convolutional Networks for Skeleton-Based Action Recognition[J].2018.
[8] Wang Q,Wu B,Zhu P,et al.ECA-Net:Efficient Channel Attention for Deep Convolutional Neural Networks[J].2019.
[9] Yang Y B.SA-Net:Shuffle Attention for Deep Convolutional Neural Networks:10.1109/ICASSP39728.2021.9414568[P].2021.
[10] SHAHOUDY A,LIU J,NG T T,et al.Ntu rgb+ d:A largescale dataset for 3d human activity analysis[C]//IEEE Conference on Computer Vision and Pattern Recognition,2016.
[11] Shahroudy A,Liu J,Ng T T,et al.NTU RGB+D:A Large Scale Dataset for 3D Human Activity Analysis[J].IEEE Computer Society,2016:1010-1019.
PDF(1306 KB)
