1.3 System Description
Our system and its computation are based on the contour of hand and fore-arm.
In our system, we assume that the input contour is the contour of hand and fore-arm.
We are not performing any hand or arm recognition. The original input data will be
the RGB image captured by a web camera. We perform some preliminary processing
to generate a binary image which provides enough information of hand contour. The
binary image will be used to calculate the contour and the convex hull of the contour.
The palm position can be initially estimated by the information we extracted from 3
convex hull, the fingertips position can be detected from the hand contour. Then the
system will make some modification of the initially estimation to make our results
more accurately. The computation cost of our system is low since only computation
geometry algorithms are performed; and there’s no need to apply classifier and no
pre-trained data base is needed.