Project overview
This project maps the role of neuromorphic vision in human–robot collaboration. It examines how event cameras can support real-time understanding of human pose, actions, gestures, facial expressions, gaze, and attention, and how those signals can be converted into safer and more responsive robot decisions.
Core contributions
- A PRISMA-guided systematic review of neuromorphic vision for human behavior understanding in robotics.
- A unified framework connecting event sensing, human-state estimation, command generation, safety assessment, and robot control.
- A structured analysis of datasets, methods, implementation barriers, and open research challenges.
- A roadmap for integrating event cameras into industrial, healthcare, assistive, and collaborative robotic systems.
Reported evidence
The review covers the progression from event-based tracking and pose estimation to action recognition, gesture understanding, facial-expression analysis, and gaze or eye tracking. It also discusses sensor integration, datasets, real-time processing, and deployment challenges.
Datasets and evaluation
- Human pose and motion-capture datasets, including event-based and hybrid event–RGB resources.
- Human action and gesture-recognition datasets for dynamic vision sensors.
- Facial-expression, gaze, eye-tracking, and human-monitoring datasets relevant to HRC.
- Comparative assessment of dataset availability, sensing conditions, task coverage, and limitations.
Relevance to human-in-the-loop robotics
The paper provides the conceptual backbone of this website. It frames event vision as a perception layer for human-aware robotics, where rapid and privacy-conscious sensing supports command interpretation, task understanding, safety assessment, and adaptive control.