An algorithm for rapid hand gesture analysis

This paper presents a new algorithm for the analysis of hand gestures. The algorithm takes as input six coordinates. These are the positions of the fingertips and wrist in 3D space. It then constructs a twenty-one point geometric model of the hand. The algorithm presented is related in principle to that developed by Lee and Kunii (1995). It uses a mathematical model of the hand structure and produces a geometric model with reference to the hands´ physiological characteristics. The major advantages of this algorithm are: the use of only six characteristic points as opposed to seven and direct model generation without iterative fitting. The six characteristic points are located by using a colour coded glove and two cameras. Results obtained using this algorithm show that very accurate geometric models of the hand can be obtained in 0.07 seconds