Advances in Human Motion Analysis and Synthesis

The field of human motion analysis and synthesis is rapidly evolving, with a focus on developing innovative methods for characterizing and predicting human movement. Recent research has explored the use of augmented reality games, footstep-induced floor vibrations, and physical plausibility-aware trajectory prediction to improve our understanding of human motion. These advances have the potential to enhance the realism of digital characters, improve gait health monitoring, and facilitate the early detection and rehabilitation of neuromusculoskeletal disorders. Noteworthy papers in this area include: ARFlow, which proposes a novel framework for human action-reaction synthesis that eliminates the need for complex conditional mechanisms. PRIMAL, which introduces an autoregressive diffusion model that generates unbounded, realistic, and controllable motion. GAITGen, which proposes a disentangled motion-pathology impaired gait generative model that generates realistic gait sequences conditioned on specified pathology severity levels.

Sources

Reaching Motion Characterization Across Childhood via Augmented Reality Games

Bridging Structural Dynamics and Biomechanics: Human Motion Estimation through Footstep-Induced Floor Vibrations

ARFlow: Human Action-Reaction Flow Matching with Physical Guidance

Physical Plausibility-aware Trajectory Prediction via Locomotion Embodiment

PRIMAL: Physically Reactive and Interactive Motor Model for Avatar Learning

AGIR: Assessing 3D Gait Impairment with Reasoning based on LLMs

GAITGen: Disentangled Motion-Pathology Impaired Gait Generative Model -- Bringing Motion Generation to the Clinical Domain

Built with on top of