【摘要】 从人手连续形变的时间模型、连续形变的时段模型、突变强度和自遮挡特征等不同角度描述人手状态变量的微观结构,将其作为设计粒子滤波跟踪算法的基本数据层,还对各帧粒子数目的动态分配方式进行了探讨.实验表明,与传统的粒子滤波算法相比较,本文基于状态变量微观结构的粒子滤波手势跟踪算法可以大幅度降低粒子数目.更多还原

【Abstract】 Used as basic data level,the microstructure of a state variable is described from the temporal mode,period mode of continuous deformation,mutation intensity and self-occlusion features,and the models are resulted from cognitive psychology analysis in the process of human-computer interaction.At the same time,the approach to dynamically assign the particles number over frames for reduction of the average particles number is probed into as well.The algorithm presented is implemented with VC++,and the experimental results are summarized and compared with other previous researches.By examining several real videos of moving hands,we find that our algorithm can decrease the sampled particle number more effectively than the classical particle filtering.更多还原