Symmetry-Driven hyper feature GCN for skeleton-based gait recognition
作者：
Highlights：
• SDHF-GCN is the first work that models the skeleton-based graph structure with graph-based neural network in gait recognition task.
• The symmetry perceptual principle enforces the dependencies between related joints and suppress the noises caused by joint estimation.
• Hierarchical features can improve the expressive and discriminative power of the gait features.
• SDHF-GCN renders substantial improvements over mainstream methods, especially in the coat-wearing scenario.
• Silhouette-based and skeleton-based action patterns are certified to be highly complementary.
摘要
•SDHF-GCN is the first work that models the skeleton-based graph structure with graph-based neural network in gait recognition task.•The symmetry perceptual principle enforces the dependencies between related joints and suppress the noises caused by joint estimation.•Hierarchical features can improve the expressive and discriminative power of the gait features.•SDHF-GCN renders substantial improvements over mainstream methods, especially in the coat-wearing scenario.•Silhouette-based and skeleton-based action patterns are certified to be highly complementary.
论文关键词：Dynamics of skeleton,Gait recognition,Graph convolutional networks,Symmetric interaction pattern,Hyper feature
论文评审过程：Received 6 November 2020, Revised 26 November 2021, Accepted 1 January 2022, Available online 8 January 2022, Version of Record 14 January 2022.
论文官网地址：https://doi.org/10.1016/j.patcog.2022.108520
