Robust fingertip tracking for constructing an intelligent room

摘要

In a watch-over system, the machine has to detect intention of human user according to circumstance, and such detection has to be able to be achieved naturally without special interface gadget like a hand controller. Hence, in this paper, we propose a hand detection and fingertip tracking method for use in an intelligent room. We are constructing an intelligent room in which any user can control home appliances with intuitive gestures. The presented intelligent room is a room in which home appliances and an interaction monitoring robot for elderly people are operated using intuitive gestures, without any additional equipment and any restriction for positions. The proposed method detects the hand region from a range data, which is robust for environmental changing in complex background. Using weak classifiers, the shape of extracted hand and direction of the fingertip is estimated without needing to know the location of the face. These processes are computationally fast enough to track an object in real time. In addition, characters written on a virtual 2D plane by the user's finger movement were estimated using incremental simple PCA for operator recognition.