基于关键特征点运动轨迹的动态手势识别

Dynamic Gesture Recognition Based on Key Feature Points Trajectory

为实现基于运动轨迹信息的动态手势识别,本文介绍了一种基于手势关键特征点轨迹识别的方法。将深度摄像机获取的深度信息经过自适应阈值算法提取人体目标,经过细化等算法得到人体骨架,并提取手势关键特征点轨迹,利用支持向量机在公开的、具有挑战性的DHA数据集中有关手势数据进行识别和评估。实验证明该方法可以实现复杂背景下的多种手势的识别,鲁棒性强。

The strategy of key feature point trajectories based gesture recognition was applied.human targets through adaptive threshold algorithm from the depth information using depth camera were extracted,and human skeleton was realized with thinning algorithm。Key feature points of the gesture trajectory were extracted。DHA dataset gesture data supporting vector machine in public was also used to identify and assess the concerning gesture.It is experimentally shown that a variety of gesture recognition in complex background with strong robustness can be achieved using this method.