{"title":"Global Structured Feature Graph Convolutional Network for Skeleton-Based Action Recognition","authors":"Chia-Fen Hsieh, Po-Jen Liao","doi":"10.1109/taai54685.2021.00026","DOIUrl":null,"url":null,"abstract":"With the development of human action recognition technology, deep learning has been applied to still images, and great progress has been made. However, in film action recognition, there is still the issue of using deep learning to improve the recognition rates. When predicting the action of a movie, encountering occlusions, large background changes, or accumulation of some errors in consecutive frames in the movie, resulting in a decrease in the accuracy of action recognition and increase the difficulty of film action recognition. In addition, there is a lack of structural information of bone joints and related research between two different structures. To solve this problem, this paper proposed a joint structure related feature network method using graph convolution network (GCN), which combines multiple convolution kernels of different dimensions to enhance the recognition rate of movie actions. The experimental database was established in the laboratory of Nanyang Technological University, Singapore. The system uses the NTU RGB+D motion recognition data set to evaluate our network. Preliminary experimental results show that our system may improve accuracy and make it more efficient.","PeriodicalId":343821,"journal":{"name":"2021 International Conference on Technologies and Applications of Artificial Intelligence (TAAI)","volume":"162 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2021-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2021 International Conference on Technologies and Applications of Artificial Intelligence (TAAI)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/taai54685.2021.00026","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0
Abstract
With the development of human action recognition technology, deep learning has been applied to still images, and great progress has been made. However, in film action recognition, there is still the issue of using deep learning to improve the recognition rates. When predicting the action of a movie, encountering occlusions, large background changes, or accumulation of some errors in consecutive frames in the movie, resulting in a decrease in the accuracy of action recognition and increase the difficulty of film action recognition. In addition, there is a lack of structural information of bone joints and related research between two different structures. To solve this problem, this paper proposed a joint structure related feature network method using graph convolution network (GCN), which combines multiple convolution kernels of different dimensions to enhance the recognition rate of movie actions. The experimental database was established in the laboratory of Nanyang Technological University, Singapore. The system uses the NTU RGB+D motion recognition data set to evaluate our network. Preliminary experimental results show that our system may improve accuracy and make it more efficient.