Research On Real-time Dynamic Hand Gesture Segmentation Method With Complex Background

Hand gesture, as an important way of Human-Computer Interaction, plays as a significant role in the field of Human-Computer Interaction. Hand gesture segmentation is the initial stage of hand gesture recognition, and the results of obtained of hand gesture segmentation will directly influence the follow-up work. Therefore, hand gesture segmentation plays a decisive role in the whole process of hand gesture recognition. This thesis is mainly based on the vision based hand gesture segmentation method, combining theory and experiment, and has studied hand gesture segmentation methods based on vision to segment hand gesture images or target hand gesture of video sequences from complex background. Specific contents are as follows:1. By studying existing commonly used color spaces such as RGB, HSV and YCb Cr, a kind of hand gesture segmentation method based on YCb Cr color space and K-means clustering is proposed in this thesis. Experiment results illustrate that the proposed method can segment hand gesture from complex backgrounds and obtain good hand gesture segmentation results. The proposed method is also robust to changing illuminance. The proposed method applys to static hand gesture segmentation with complex background and illumination interference.2. Aiming at the disadvantages of existing skin color model, on the basis of research above, a kind of hand gesture segmentation based mixed skin-color model and FCM algorithm is proposed. The experiment results illustrate that the proposed method is simple, efficient and easy to implement. And it can be used in the field of dynamic hand recognition with complex backgrounds. The proposed method applys to static hand gesture segmentation with complex background and nature enrivenment interference.3. To solve problems of segmentation accuracy and real-time that exist in motion hand gesture with complex backgrounds,a kind of dynamic hand gesture segmentation method based on improved Kalman filter and weighted skin-color model is proposed. The experimental results illustrate that utilizing the proposed method can segment hand gesture from dynamic hand gesture video sequences with complex backgrounds effectively. And the accuracy of hand gesture segmentation is high. The proposed method applys to dynamic hand gesture segmentation of video sequence with complex color and illumination interference.4. Aiming at the disadvantages of existing hand gesture segmentation methods, a novel dynamic hand gesture segmentation method which combines unequalprobabilities and improved Fuzzy C-Means algorithm is proposed. Firstly, this method utilizes unequal-probabilities to build background model of complex dynamic background and detects motion region of hand gesture with background difference. Secondly, FCM algorithm is improved to accelerate the rate of convergence, cluster hand gesture image and distinguish skin-color region and non-skin-color region. The experiment results illustrate that the proposed method has high accuracy and is robust to skin-color interference and varying illumination. The proposed method applys to dynamic hand gesture segmentation of video sequence with n on target hand, face and illumination interference.In consideration of the practical application meaning of gesture recognition in human-computer interaction, improving the accuracy of dynamic gesture segmentation and real-time of segmentation algorithm in complex background needs further studied.