首页 | 本学科首页   官方微博 | 高级检索  
     检索      

基于光流及耦合隐马尔可夫模型的动态手势识别
引用本文:刘江华,陈佳品,程君实.基于光流及耦合隐马尔可夫模型的动态手势识别[J].上海交通大学学报,2003,37(5):720-723,736.
作者姓名:刘江华  陈佳品  程君实
作者单位:上海交通大学信息存储研究中心,上海,200030
摘    要:基于块的相关算法来计算光流,并利用光流跟踪双手的运动.双手的运动轨迹取相邻两点的速度向量,经8方向链码量化后作为观察向量.和直接利用位置信息相比较,提高了识别的鲁棒性.采用耦合隐马尔可夫模型来识别双手动态手势,提出并实现了最大后验概率的训练.对6个双手动态手势的试验表明,耦合隐马尔可夫模型(CHMM)比常规隐马尔可夫模型(HMM)能更有效地对双手动态手势建模.

关 键 词:耦合隐马尔可夫模型  最大后验概率  动态手势识别  光流跟踪
文章编号:1006-2467(2003)05-0720-04

Dynamic Hand Gesture Recognition Based on Optical Flow and Coupled Hidden Markov Model
LIU Jiang hua,CHEN Jia pin,CHENG Jun shi.Dynamic Hand Gesture Recognition Based on Optical Flow and Coupled Hidden Markov Model[J].Journal of Shanghai Jiaotong University,2003,37(5):720-723,736.
Authors:LIU Jiang hua  CHEN Jia pin  CHENG Jun shi
Abstract:Optical flow was computed based on the block relative algorithm, which is used for hand tracking. Velocity vector between two adjacent hand positions is extracted and quantified by eight directional chain code as observation vector, which provides more robust recognition rate than position vector. CHMM (Coupled Hidden Markov Model) was used to recognize dynamic two hand gestures. MAP (Maximum a posteriori) was proposed and realized for CHMM's training. Experiments on six dynamic two hand gestures show that CHMM is more effective to model two hand gesture than traditional HMM.
Keywords:coupled hidden Markov model (CHMM)  maximum a posteriori (MAP)  dynamic gesture recognition  optical tracking
本文献已被 CNKI 维普 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