基于改进型形状上下文描述子的字母手势识别

针对基于单目视觉信息的裸手手势,采用了基于改进型形状上下文描述子的分类识别方法。该方法首先通过肤色信息以及背景建模提取手部区域,然后利用单手指模板对手指进行检测,同时采用改进型形状上下文描述子对手部区域整体轮廓进行描述。在此基础上,使用有向无环图支持向量机(Directed Acyclic Graph Support Vector Machine,DAGSVM)对所提取的特征进行模式分类。其中,针对基本算法存在的问题,改进型形状上下文描述子将基于各个轮廓点的形状上下文直方图改为基于重心的形状上下文直方图,以提高计算速度,增强实时性。对30种字母手势,3种控制手势和10个数字手势开展的离线和在线实验结果表明,该方法取得了较好的分类准确率(离线:96%,在线:91%)和较高的实时性(识别时间14~15ms),适用于基于字母手势的实时人机交互。

Alphabetic Gesture Recognition Based on the Improved Shape Context Descriptor

The classification identification method based on improved shape context descriptor is adopted to classify the bare-hand alphabetic gestures based on monocular vision. In this method, the skin color information and background model are employed to extract the hand region, and single finger template is utilized for the finger detection. Then the overall contour of hand region is described by the improved shape context descriptor, and the DAGSVM is used as classifier. To decrease the computing time-consumptio...