Artificial neural networks for real-time optical hand posture recognition using a color-coded glove

Abstract

Optical pose recognition of the hand is an extremely attractive method for user-computer interaction in many applications. The image of a hand in the frame of a video camera is processed and the pose it is making, its current finger configuration, is detected. Often combined with position tracking, it allows for a very natural way of giving commands. Furthermore, it alleviates the use of sometimes cumbersome pieces of hardware. Within immersive virtual reality systems, the liberty of movement of the commanding hand requires extra considerations not normally dealt with by typical optical hand posture recognition interfaces for desktop system applications. This research proposes an artificial neural network approach to the recognition of hand postures. The optical capture inside an immersive virtual reality workspace and the extraction of features of this hand are facilitated by the use of a specially coded color glove.