摘要
为解决复杂背景下手势图像不能准确分割的问题,结合YCg Cr颜色空间上的聚簇特性,提出一种在RGB与YCg Cr双颜色空间下的肤色检测法。该方法有效地减少了背景中类肤色的干扰,完成了手部图像在复杂背景下的检测和分割;并采用HOG特征对手势图像进行表示,最后采用直方图相交核支持向量机进行手势识别。该方法在塞巴斯蒂安.马塞尔静态手姿数据库上进行了测试,对6种手势的平均识别率达到了93.33%。
Accurate segmentation of hand gesture is difficult to achieve under complex background. Combining the poly cluster characteristics of YCgCr color space, a novel color detection method is proposed under double color space of RGB and YCgCr. This method can effectively reduce the interference of similar skin color in background, and complete the detection and segmentation of the hand images in complex background. HOG features is chosen to represent the hand gesture image. Hand gesture recognition is achieved by using histogram intersection nuclear based support vector machine （SVM）. Tested on Sebastian Marc＇s static hand gesture databases, the proposed method can achieve average 93.33% recognition rate for six hand gestures.
出处
《自动化技术与应用》
2016年第8期113-116,共4页
Techniques of Automation and Applications
基金
国家自然科学基金项目资助(编号61103123
61203021)
关键词
YCGCR颜色空间
HOG特征
直方图相交核
支持向量机
手势识别
YCgCr color space
HOG features
histogram intersection kernel
support vector machine （SVM）
gesture recognition
