A REVIEW OF HUMAN SKELETON ACTION RECOGNITION METHODS BASED ON GRAPH CONVOLUTIONAL NETWORK
Skeleton-based action recognition has emerged as one of the most prominent and crucial research areas in computer vision.The use of human skeleton data in action recognition provides a higher degree of robustness to changes in light,background,and perspective,compared to other data modalities.Human skeleton data exists in the form of a topological graph structure,and graph convolution,which is a deep learning method based on graph structures,is capable of extracting and classifying the features of human skeleton data in an efficient manner.Therefore,graph convolution-based methods have become the mainstream approach for processing skeleton data.It is essential to comprehensively and systematically summarize and analyze the action recognition methods based on graph convolution.Accordingly,this paper provides a comprehensive review of the latest advancements in action recognition technology based on graph convolution methods.The relevant methods are classified and summarized,and the benchmark dataset is studied in detail.Finally,the future research direction and trend are discussed.
万方数据
维普
