基于图卷积神经网络的人体骨架动作识别研究进展  

Review of Human Skeleton Action Recognition Based on Graph Convolution Neural Network

作　　者：杨巨成[1] 张泉钰 王波 王嫄 陈亚瑞[1] 赵婷婷[1] YANG Jucheng;ZHANG Quanyu;WANG Bo;WANG Yuan;CHEN Yarui;ZHAO Tingting(College of Artificial Intelligence,Tianjin University of Science&Technology,Tianjin 300457,China;Sitonholy(Tianjin)Technology Co.,Ltd.,Tianjin 301799,China)

机构地区：[1]天津科技大学人工智能学院,天津300457 [2]思腾合力(天津)科技有限公司,天津301799

出　　处：《天津科技大学学报》2023年第6期1-11,共11页Journal of Tianjin University of Science & Technology

基　　金：国家自然科学基金项目(61976156)。

摘　　要：基于人体骨架的动作识别是实现计算机视觉智能的重要分支。本文对基于图卷积神经网络的人体骨架动作识别技术进行研究并分析,对基于频谱图卷积和空域图卷积的研究现状进行综述,并从邻接矩阵和输入特征两个角度详述了图卷积模型在人体骨架动作识别领域的研究进展。此外,对现有的基于图卷积神经网络的人体骨架动作识别算法进行了分析比较,最后展望了图卷积神经网络在人体骨架动作识别领域的未来发展方向。Action recognition based on human skeleton is an important branch to achieve computer vision intelligence.In this article we first review the research on human skeleton action recognition based on graph convolutional neural networks and analyze the key techniques.Then,we outline the research progress of graph convolution approaches based on spectral convolution and space domain convolution,and detail the research progress of graph convolution models in the field of human skeleton action recognition from two perspectives of adjacency matrix and input features.Furthermore,we analyze and compare the existing algorithms for human skeleton action recognition based on graph convolution neural networks.Finally,we look forward to the future development direction of graph convolution neural networks in the field of human skeleton action recognition.

关 键 词：图理论 图神经网络 图卷积神经网络 基于骨架的动作识别 时空域融合 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]