Stereo Vision-based Hand Tracking And Recognition In The Human-computer Interaction

Nowadays human-computer interaction technology becomes a new hot spot of computer research field. Human-computer interaction is an interface of information exchanges between human and computer systems through the input and output devices of systems. Traditional human-computer interaction has been unable to meet the needs of the natural and diversity. As computer technology continues to progress, human-computer interaction will gradually achieve a seamless connection with the direction of humane and intelligent.Because of the agility and breadth of hand gestures in interpersonal communi-cation, this paper selects the hand gestures as the interface of human-computer inter-action. In this paper, the stereo camera Bumblebee2is chose to collect RGB image and depth information. Based on this information, we propose a new simple gestures3D tracking and recognition algorithm, and present an application in human-computer interaction.In complex background, convert the RGB image to YCbCr space. Interested ges-ture region is obtained through the OSTU adaptive threshold segmentation and the maximum connected domain selected algorithms. Using distance transform to extract center of palm, and get depth information through the stereo matching. Here the three-dimensional coordinates of the palm’ center can be used for3D hand tracking. At the same time, calculate curvatures on the contour and fit ellipses to identify fingertips, as a simple gesture recognition using the numbers of fingertips. Compared to tradi-tional hand gesture tracking and recognition schemes, the method this paper proposed can work well and stability in the case of more complex background and the impact of illumination. The restrictions on users of our system is relatively small. The most important thing is that we can do hand tracking and recognition in three-dimensional space. And this3D information can be applied to many human-computer interaction occasions.