Hand image interpretation based on double active contour tracking

We propose an approach to hand sign interpretation in image that is based on active contour tracking. We can decompose our approach into 5 steps: a color-based skin pixel detection, a double hand contour detection, the localization of fingers and palm (the hand description generation), the detection of a final position (with respect to considered signs) and finally, the interpretation of a single position or a sequence of positions in terms of a hand sign. We employ a double active contour-based finger and palm localization in the image and a subsequent interpretation in terms of signs. As a final result, 21 different signs are recognized, that correspond to hand positions (i.e. the visibility of palm, fingers and thumb).