基于GCN的局部增强人体骨骼行为识别算法  

Locally enhanced human skeleton action recognition algorithm based on GCN

作　　者：谢伙生[1] 罗洪文 XIE Huo-sheng;LUO Hong-wen(College of Mathematics and Computer Science,Fuzhou University,Fuzhou 350591,China)

机构地区：[1]福州大学数学与计算机科学学院,福建福州350591

出　　处：《计算机工程与设计》2022年第6期1777-1784,共8页Computer Engineering and Design

基　　金：国家自然科学基金项目(61801121)。

摘　　要：为使人体骨骼这种非欧几里得结构数据更好地应用于人体行为识别任务中,将多种模态的特征融合后作为模型的输入,探讨关节的全局与局部的信息共现性,构建一个完整的空间信息推理模型。通过使用多层时间膨胀卷积网络提取骨骼数据在时间维度中的上下文依赖关系,使模型可以兼具捕获短期和长期的时序上下文依赖关系,有效提高时间信息的提取能力。提出的分层次时空图卷积神经网络模型在NTU-RGB+D-60和NTU-RGB+D-120数据集上均取得了较好的识别准确率。To better apply non-Euclidean structural data like human skeleton to human action recognition tasks,the features of multiple modes were taken as input to the model,global and local information co-occurrence of joints was explored,and a complete spatial information reasoning model was constructed.A multilayer temporal dilation convolution network was used to extract the context dependence of skeleton data in the temporal dimension,which enabled the model to capture both short-term and long-term time series context dependencies,and effectively improves the ability to extract temporal information.The hiera-rchical spatial-temporal graph convolution neural network model proposed achieves good recognition accuracy on NTU-RGB+D-60 and NTU-RGB+D-120 datasets.

关 键 词：图卷积神经网络 人体骨骼数据 行为识别 膨胀卷积 时间卷积网络 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]