南京师大学报（自然科学版）2025，Vol.48Issue(3)：112-119,8.DOI:10.3969/j.issn.1001-4616.2025.03.013
基于多流自适应时空图卷积网络的人体行为识别
Human Behavior Recognition Based on Multi-stream Adaptive Spatio-temporal Graph Convolutional Networks
摘要
Abstract
Aiming the problems of insufficient mining of temporal and spatial features in current human action recognition algorithms,a human skeleton action recognition model based on multi-stream adaptive spatio-temporal graph convolutional networks is proposed in this paper.Firstly,the attention mechanism and neural tensor network(NTN)algorithm are used to solve the connection strength between each pair of nodes,and the global adjacency matrix is constructed.Then,topK strategy is used to dynamically select the first K neighbor nodes according to the connection strength and update the global adjacency matrix.Next a hybrid pooling model is used to extract the global context information and time key frame features.By modeling joint information,bone information,joint movement information and bone movement information,the ability of the features extracted from the model to represent the movement is strengthened.Finally,experiments are carried out on the data set NTU-RGB+D.The results show that the model has a good performance in the task of human skeleton action recognition,and effectively improves the accuracy of action recognition.关键词
动作识别/注意力机制/图卷积神经网络/多流网络Key words
action recognition/attention mechanism/graph convolutional neural network/multistream network分类
信息技术与安全科学引用本文复制引用
古学茹,张备伟,丁雯..基于多流自适应时空图卷积网络的人体行为识别[J].南京师大学报（自然科学版）,2025,48(3):112-119,8.基金项目
国家自然科学基金项目(61877061). （61877061）
