连续动态手势的时空表观建模及识别

论述了复杂背景下连接动态手势的分割、建模及识别；融合手势运动信息和皮肤颜色信息，进行复杂背景下的手势分割；通过结合手势的时序信息、运动表观以及形状表观，提出动态手势的时空表观模型，并提出基于颜色、运动以及形状等多模式信息的分层次融合策略抽取时空表观模型的参数，最后，提出动态时空规整算法用于手势识别，实验表明，利用上述提出的手势分割、建模、特征参数抽取及识别方法识别１２种手势，平均识别率高达９７％。

Spatio-Temporal Appearance Modeling and Recognition of Continuous Dynamic Hand Gestures

This paper presents the segmentation, spatio temporal appearance modeling and recognition of continuous dynamic hand gestures with complex background. Fusing the coarse motion information and skin color chrominance, the segmentation of hand gesture can be got robustly. By integrating temporal information, motion appearance, and shape appearance, a spatio temporal appearance model is proposed to represent dynamic hand gestures, and a novel scheme is proposed for recovering the model parameters by hierarchical fusion of multi modal information such as skin chrominance, motion, and shape. In addition, Dynamic Space Time Warping is suggested for recognition of hand gestures. Experiment results indicate that the average recognition rate of the 12 hand gestures is 97% based on the proposed segmentation, modeling, feature extracting, and recognition techniques.