Agentic data science with the Data Agent Kit
cloud.google.com·7h ago
AI Research
arXiv CS.AI·Sep 9·5 min read
Subject-Relative Micro-Motion and Sleep Dynamics for Near-Infrared Video Sleep Staging
TL;DR
ViNUSS, a novel framework for sleep staging using near-infrared video, achieves 0.80 accuracy and 0.78 macro-F1 score by analyzing micro-motions and sleep dynamics without relying on physiological proxies.
✦ Why It Matters
Consider implementing ViNUSS for enhanced sleep monitoring in your research or clinical practice.
Key Takeaways
- Explore ViNUSS for video-based sleep analysis.
- Utilize NIR video to avoid physiological signal dependency.
- Focus on micro-motion patterns for sleep stage cues.
- Analyze thoraco-abdominal movements for better interpretability.
How It Works
ViNUSS combines subject-relative micro-motion learning with full-night sleep dynamics modeling. It encodes micro-motions while preserving spatial context, allowing the model to learn sleep stage cues based on individual sleep patterns.
The two-scale modeling captures motion changes within sleep epochs and organizes evidence into a comprehensive sleep-stage trajectory.