中文摘要：

提出了一种基于运动变化的实时指势判别方法。根据指势行为特征，跟踪手臂运动区域的运动状态，计算指势视频中每个独立运动序列的连续运动方向，统计其在一定方向范围内的累积运动帧数，当该运动序列在确定方向范围内的累积运动帧数达到预定范围时，根据指势手臂在指向目标时存在片刻停顿，从复杂场景中实时判别指势。实验结果表明，该方法简单、有效。

英文摘要：

A real-time pointing gesture discrimination based on motion variation is developed in this paper. According to the characteristics of the pointing, the motion states of every arm motion region are tracked. The continuous motion orientations of every separate motion sequence from a pointing video are calculated and the accumulative frames in a certain direction are counted. When the accumulative frame number in a certain direction from the motion sequence reaches a scheduled range, the pointing is fast discriminated from a clutter scene according to the knowledge that there is a pause while pointing. Experimental results show that the method is simple and efficient.