Real-time Hands Tracking Using Feature Point Gathering Based on KLT Tracker for Man-Machine Interface

摘要

@@ Intuitive man-machine interfaces based on gestures with a touchpad device have become common. Furthermore, a vision based gesture recognition system like Kinect is gradually spread. Conventional works, however, use complex input devices (plural cameras, sensors, and so forth) or need to wear some devices like hand globes that is limitation for manmachine interface. This paper proposes a real-time single-input both hands tracking algorithm for intuitive man-machine interfaces. By applying feature-point gatherings into the KLT (Kanade- Lucas-Tomasi) tracker, a kind of an optical flow, non-rigid objects like hands can be traced with high accuracy and low complexity under a complex background.