Hand Gesture Recognition Based On Kinect And Three-finger Dexterous Hand Interaction

With the rapid development of computer technology,the development of human-computer interaction technology is a great advance.People are no longer satisfied with the traditional contact mode,they want a more natural,non-contact human-computer interaction.In the natural human-computer interaction,the main research contents are gesture recognition,speech recognition and face recognition.Gestures are the most natural and main interactive part of human beings.They attract more attention in these technologies.Vision-based gesture recognition technology has become a research hotspot in the field of gesture recognition,but the traditional vision based gesture recognition technique is usually based on skin color model,which is easily influenced by the illumination,environment,the similar skin color and complex background.The introduction of the Kinect and other depth cameras bring new solutions to these problems.I read a lot of domestic and foreign literature.By comparing these researches and combining current issues,I designed a method of hand gesture recognition based on depth-image segmentation,and a stone-scissors-cloth game with three finger dexterous hands for human-machine interaction.This method first obtains the depth ima ge through the Kinect depth camera,then uses the Open NI and the Ni TE to obtain the the rough position of the hand,according to the hand position distance value carries on the depth threshold segmentation,simultaneously tracks the hand,and compares with the skin color model.Then the edge of the image is detected by the canny operator,the external contour of the hand is extracted.The center moment of the outer contour of the gesture is computed,and the fingertip coordinates are obtained by improving t he traditional barycenter-distance method.Gesture recognition classification based on the coordinates of fingertips.Using this improved gesture recognition classification method,we tested the common 8 gestures.The robustness of the algorithm is validat ed under various conditions,and the improved method is also tested in the color space and has a great result.Then the extra three interactive gestures "stone","scissors" and "cloth" are tested again.Finally completes the human-machine interaction stone-scissors-cloth game.Programming the three-fingered dexterous hand to complete a human-machine interaction by the designed strategies.Finally verifies the human-machine interaction result.This method can solve the problem of skin segmentation easily affected by environment,light,skin color and other complex background.This method do not need collect a large number of templates and do not need to carry out complex model training.It has a high flexibility.And the features are simple but efficient,so the recognition speed is fast and the recognition rate is high.The whole system has good real-time performance and high robustness.