Spatiotemporal Graph Autoencoder Network for Skeleton-Based Human Action Recognition

Human action recognition (HAR) based on skeleton data is a challenging yet crucial task due to its wide-ranging applications, including patient monitoring, security surveillance, and human- machine interaction. Although numerous algorithms have been proposed to distinguish between various activities, most practical applications require highly accurate detection of specific actions. In this study, we propose a novel, highly accurate spatiotemporal graph autoencoder network for HAR, designated as GA-GCN. Furthermore, an extensive investigation was conducted employing diverse modalities. To this end, a spatiotemporal graph autoencoder was constructed to automatically learn both spatial and temporal patterns from skeleton data. The proposed method achieved accuracies of 92.3% and 96.8% on the NTU RGB+D dataset for cross-subject and cross-view evaluations, respectively. On the more challenging NTU RGB+D 120 dataset, GA-GCN attained accuracies of 88.8% and 90.4% for cross-subject and cross-set evaluations. Overall, our model outperforms the majority of the existing state-of-the-art methods on these common benchmark datasets.