首页 | 本学科首页   官方微博 | 高级检索  
     

线上线下结合的实时手势跟踪系统
引用本文:王超,崔更申,黄廷辉,余益民.线上线下结合的实时手势跟踪系统[J].桂林电子科技大学学报,2012,32(2):125-128.
作者姓名:王超  崔更申  黄廷辉  余益民
作者单位:桂林电子科技大学 计算机科学与工程学院,广西 桂林,541004
摘    要:手势识别跟踪一般采用线下训练分类器,不能有效检测跟踪形变的手势,针对手势形变及在窗口的暂时性消失等问题,提出了一种通过线下训练结合线上提取样本对分类器进行训练的检测方法,同时采用跟踪一检测一学习(TLD)的方法不断对跟踪器的结果进行纠正.试验结果表明,本算法对手势形变、短暂消失具有很好的适应性,与TLD算法相比较具有更好的稳定性.

关 键 词:手势识别  手势跟踪  Haar特征级联器  线上检测器

Hand gesture tracking system based on online and offline
Wang Chao , Cui Gengshen , Huang Tinghui , Yu Yimin.Hand gesture tracking system based on online and offline[J].Journal of Guilin Institute of Electronic Technology,2012,32(2):125-128.
Authors:Wang Chao  Cui Gengshen  Huang Tinghui  Yu Yimin
Affiliation:(School of Computer Science and Engineering,Guilin University of Electronic Technology,Guilin 541004,China)
Abstract:Hand gesture recognition and tracking methods usually adopt offline training classifier,but the deformation of hand gesture can cause drift problem.In order to achieve more stability long time tracking,a novel detection method based on offline and online was presented,combined with TLD method to revise tracker.Experimental result shows this system could achieve more stability in hand gesture deformation and provisional disappear and it is more stable than TLD algorithm.
Keywords:hand gesture recognition  hand gesture tracking  Haar cascade classifier  online classifier
本文献已被 CNKI 万方数据 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号