Gesture segmentation based on monocular vision using skin color and motion cues

Abstract

Gesture segmentation is the first and the most critical step in sign language recognition. In this paper a method of gesture segmentation from the video image sequence based on monocular vision is presented by skin color and motion cues. Firstly, determine the background images and capture the gesture images with a sampling interval of 10 frames. Then with the difference method, the movement region of hand gestures is detected, meanwhile, through the analysis of the skin color information, the skin color region of hand gestures is obtained. After that, the initial gesture region can be gained by the AND operation on the movement region and the skin color region. Last but not least, gestures are separated from video image sequence reliably and completely using the mathematical morphology method. The experimental results indicate that the technique is capable of segmenting the gestures quite effectively.