GeomNet: A Neural Network Based on Riemannian Geometries of SPD Matrix\n Space and Cholesky Space for 3D Skeleton-Based Interaction Recognition

2021/11/25 by Nguyen, Xuan Son · 2 citations
Computer Science · Engineering · #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Gait Recognition and Analysis #Hand Gesture Recognition Systems #Human Pose and Action Recognition #Machine Learning (cs.LG) #Machine Learning (stat.ML)

paper · pdf · doi:10.48550/arxiv.2111.13089

openalex publication_date 2021/11/25 · openalex created_date 2022/11/06 · openalex updated_date 2026/07/28

Abstract

In this paper, we propose a novel method for representation and\nclassification of two-person interactions from 3D skeleton sequences. The key\nidea of our approach is to use Gaussian distributions to capture statistics on\nR n and those on the space of symmetric positive definite (SPD) matrices. The\nmain challenge is how to parametrize those distributions. Towards this end, we\ndevelop methods for embedding Gaussian distributions in matrix groups based on\nthe theory of Lie groups and Riemannian symmetric spaces. Our method relies on\nthe Riemannian geometry of the underlying manifolds and has the advantage of\nencoding high-order statistics from 3D joint positions. We show that the\nproposed method achieves competitive results in two-person interaction\nrecognition on three benchmarks for 3D human activity understanding.\n

Citations

• Co-occurrence Feature Learning for Skeleton based Action Recognition using Regularized Deep LSTM Networks
• Modeling Temporal Dynamics and Spatial Configurations of Actions Using Two-Stream Recurrent Neural Networks

Cited by

• DFNN: A Deep Fréchet Neural Network Framework for Learning Metric-Space-Valued Responses
• Learning Adaptive Node Selection with External Attention for Human Interaction Recognition

Related

• Building Neural Networks on Matrix Manifolds: A Gyrovector Space Approach
• Riemannian Deep Learning: Modules, Networks, and Geometries
• Riemannian Local Mechanism for SPD Neural Networks
• A Novel Geometric Framework on Gram Matrix Trajectories for Human Behavior Understanding
• A CNN for homogneous Riemannian manifolds with applications to Neuroimaging