Event-based tracking of human hands

This paper proposes a novel method for human hands tracking using data from an event camera. The camera detects changes in brightness, measuring motion, with low latency, no motion blur, power consumption and high dynamic range. Captured frames are analysed lightweight algorithms reporting 3D hand position data. chosen pick-and-place scenario serves as example input collaborative human-robot interactions obstacle avoidance safety applications. Events pre-processed into intensity frames. regions of interest (ROI) defined through object edge activity, reducing noise. ROI features extracted use in-depth perception. Event-based demonstrated feasible, real time at computational cost. proposed ROI-finding reduces noise images, achieving up to 89% reduction relation the original, while preserving features. depth estimation error ground truth (measured wearables), measured warping single camera, is 15 30 millimetres, depending on plane it measured. Tracking space define (hands space).