Skeleton Based Gait Recognition

Skeleton-based gait recognition aims to identify individuals based on their unique walking patterns, using only the skeletal data extracted from video. Current research focuses on developing robust and accurate models, employing architectures like transformers, graph convolutional networks, and spatial transformer networks to effectively capture both spatial and temporal aspects of gait. This field is significant because it offers a privacy-preserving biometric method less susceptible to clothing or carrying changes compared to appearance-based approaches, with applications in security and healthcare. However, ongoing work addresses the challenge of spurious correlations arising from anthropometric biases in existing datasets, highlighting the need for more diverse and realistic data for improved model generalization.