Research map / 3D vision / Human pose estimation
Optimal-state Dynamics Estimation for Physics-based Human Motion Capture from Videos
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes a monocular human motion capture method that predicts exterior forces and internal torques via a learnable PD controller driving a physics simulation, refined with a neural Kalman filter to correct simulation errors.
Problem
Monocular human motion capture yields accurate kinematics but ignores underlying dynamics (external forces and internal torques), and physics simulations introduce inaccuracies, leading to jittery or physically implausible motion.
Approach
Combine a learnable PD-controller-driven physics simulation (predicting forces and torques) with a neural Kalman filter that filters the simulated states to correct simulation inaccuracies, yielding physically plausible motion and dynamics.
Key results
- Abstract reports the approach performs well on traditional human pose estimation and that the predicted forces and torques are physically plausible compared to biomechanics literature; no specific numbers available in the provided text.
Limitations
- Physical simulation involves simplifications of real-world dynamics that introduce inaccuracies, which the filter is meant to alleviate.
Experimental setup
Builds on: neural Kalman filter PD control
Closest papers by meaning
- PhysPT: Physics-aware Pretrained Transformer for Estimating Human Dynamics from Monocular Videos (CVPR 2024 · similarity 0.88)
- MotionPRO: Exploring the Role of Pressure in Human MoCap and Beyond (CVPR 2025 · similarity 0.87)
- Dynamic Inertial Poser (DynaIP): Part-Based Motion Dynamics Learning for Enhanced Human Pose Estimation with Sparse Inertial Sensors (CVPR 2024 · similarity 0.87)
- ImDy: Human Inverse Dynamics from Imitated Observations (ICLR 2025 · similarity 0.87)
- Existence Is Chaos: Enhancing 3D Human Motion Prediction with Uncertainty Consideration (AAAI 2024 · similarity 0.87)
- InterMimic: Towards Universal Whole-Body Control for Physics-Based Human-Object Interactions (CVPR 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.