基于属性计算网络的动态手势识别的研究

Dynamic Hand Gesture Recognition based on Attribute Computing Network

基于视觉的动态手势识别既是当今人机交互系统研究的热点也是难点。该文提出了基于属性计算网络的识别方法对动态手势进行识别,与现有的识别方法相比,具有明显的特点和优点。首先利用肤色分离手部对象,然后跟踪手部对象,并对手部对象轮廓和运动轨迹抽取特征并建立对应的特征向量,最后将手部对象轮廓特征和运动轨迹特征进行加权组合,得出识别结果。本方法的特点和优点是识别分析速度较快,能够很好的满足人机交互系统实时性的需要,并且能够适应动态手势表达在速度和幅度上的差异性。

Vision-based dynamic gesture recognition system for human-computer interaction is the current hotspot and is also a difficult spot. In this paper,attribute-based computing networks for dynamic gesture recognition method to identify,as compared with the existing method of identification,has distinctive features and benefits. First we use color to get the separation of the hand object,and then track the hand object and the object contours and movement of hand trajectory extract features. Then build the corresponding feature vectors,and finally the hand object contour features and characteristics of trajectories weighted combinated and the recognition results obtained . The characteristics and advantages of this method is to identify fast and the analysis speed can meet the human-machine interaction system＇s needs of real-time,and be able to adapt the dynamic hand gestures＇s expression with speed and magnitude differences.