Research On Hand Gesture Recognition Method Based On Video Image

With the development of human-computer interaction technology, as the the most natural and common human interaction, hand gesture recognition technology has become one of the most focused research in the field of human-computer interaction. This article stand from current deficiency of the difficulty to eliminate facial interference in hand gesture recognition technology. So in order to enable it to adapt to more common application environment, this paper proposes an algorithm which combined with and based on various basic algorithm hand gesture segmentation method, realized people most commonly used gestures recognition and real-time matching template display.Segmentation algorithm in this paper, first of all, using dynamic background difference. This algorithm effectively uses the relevant information of adjacent frames in the video, considering face as part of the background and updated it constantly, and thus implemented initial segmentation under the disturbance of skin color and face. Then, in order to get more accurate hand’s outline, we introduced the watershed algorithm. This algorithm takes advantage of the topographical principle so as to realize better edge segmentation. And in order to remove the effects of part of arm, we combined with dynamic skin color segmentation which selected the real-time images of human skin color as standard and set the upper and lower threshold value range, so as to realize the skin color segmentation more accurately. Immediately, we do bitwise operation of the two values and then obtain accurate hand-shape contour. Finally, we do median filter with the binary figure to reduce serrated outline and make the hand edge more smooth.Gesture recognition algorithm in this paper is combined with Hu moment and fingertips identify to get accurate static hand gestures’ recognition. What’s more, we stand from the habit and speed of hand’s movement and divided gestures into two situations that include slow motion (including static situation) and rapid movement, then we use different way to recognize gesture and matching template based on two situations. It make the gesture recognition’s process more intuitive and also achieves good accuracy.