基于用户行为分析的手势跟踪方法  

Hand Tracking Based on Behavioral Analysis for Users

作　　者：冯志全[1,2] 杨波[1,2] 郑艳伟[1,2] 徐涛[1,2] 唐好魁[1,2] 

机构地区：[1]济南大学信息科学与工程学院,山东济南250022 [2]山东省网络环境智能计算技术重点实验室,山东济南250022

出　　处：《软件学报》2013年第9期2101-2116,共16页Journal of Software

基　　金：国家自然科学基金(61173079;60973093;61173078);新世纪优秀人才支持计划(NCET-0863);山东省自然科学基金(ZR2011FZ003)

摘　　要：人手结构的高维性而导致粒子滤波跟踪方法中采样数目非常庞大,是实现运动人手的实时性跟踪研究的主要障碍之一.以降低粒子数目为目标,以行为分析和建模为切入点,提出一种手势跟踪方法.首先分析操作者在手势操作过程中的行为特征,建立人手运动的动态模型;其次,研究动态模型的基本特征,并给出一种描述方法;然后,建立人手运动的时段模型,分析了手势状态的时间-空间关系.在此基础上,提出了状态变量微观结构的概念,重点给出了基于状态变量微观结构的手势跟踪算法;最后,设计和完成了实验,并与相关参考文献方法的实验结果进行对比.结果表明,采用该算法,用少量粒子就可以得到比较精确的跟踪结果.提出的核心算法已经用于一个基于自然手势交互的三维虚拟装配原型系统.On the way forward to real time in the process of particle-filtering-based human hand tracking, one of the main obstacles is to generate a great deal of particles which are derived from high dimensionality of human hand model. Aimed at reducing the particle number, a new particle filtering approach is put forward in this paper. First, the operator＇s cognitive psychology features in the process of human-computer interaction are analyzed and studied and a general dynamic motion model is constituted. Second, some basic features of the dynamic motion model are studied and mathematically described. Third, a zonetime model of the moving human hand is proposed, and furthermore, the features in time-space of a hand gesture state are discussed. Based upon the abovementioned job, a new concept, microstructure of state variable, is presented, upon which the novel hand gesture tracking algorithm is put forward. Finally, experiments are implemented including some comparison experiments, and the algorithm is also compared with some referenced algorithms. The main contribution is that the study describes and models hand gesture behaviors and connect them with freehand tracking. The experimental results show that just using a small quantity of particles, compared with the referenced algorithms, the algorithm can obtain satisfactory results.

关 键 词：人手跟踪 粒子滤波 状态变量微观结构 降低粒子数目 行为特征 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]