Human Movement Foundation Models

Movement foundation models seek transferable representations of human motion across sport, rehabilitation, robotics and daily life.

Movement knowledge remains fragmented across tools and disciplines. Shared representations could expand access to useful assessment and coaching.

Video models, wearable sensors and scalable pose data are reducing the cost of learning from movement outside specialist laboratories.

Markerless sensing and video-native models make population-scale movement data more accessible.

Motion analysis is moving from task-specific rules toward representations learned across bodies, environments and objectives.

• Vicon
• SWORD Health
• WHOOP

• Which representations transfer across populations?
• How should physical safety constrain generated guidance?

• Dataset bias
• Weak ground truth outside labs
• Limited longitudinal validation

• Accessible movement assessment
• Adaptive coaching systems
• Cross-domain motion infrastructure

• Video-first functional screening
• Transferable movement embeddings
• Human-in-the-loop coaching systems

Read the Qoriant Research Standard ↗