Gesture segmentation based on neural net and motion information

In this paper, a method combining skin color with motion information to detect and segment gesture in complex background is proposed. We put in RGB information of a pixel, judge the output and obtain the information that belongs to the skin area. Suppose that the hand motion is the master motion and by twice clustering, we can localize the hand. Experiment result shows this method is adaptive and robust, and can be adopted in many other research regions.