姿势内核学习融合决策森林在线手势识别算法

Online Gesture Recognition Algorithm Based on Fusion of PKL and Decision Forests

针对在线姿势识别中来自流行深度传感器的噪声数据影响识别鲁棒性的问题,提出了一种基于姿势内核学习融合决策森林方法。首先,将使用骨架关节角表示每种姿势;然后,利用多类SVM分类器获得姿势内核;最后,利用决策森林实时标记关键姿势序列,根据关键姿势序列完成识别。实验结果表明,本方法的识别率可高达99.3%,相比几种较为先进的识别方法,本文方法具有更好的识别鲁棒性,并且在一定程度上降低了识别所耗时间。

For the issue that noisy data from popular depth senso∽ will impact recognition robustness in online gesture recognition, a method based on fusion of pose kernel learning and decision forests is proposed. Firstly, each pose is described using an angular representation of the skeleton joints. Then, SVM classifier with multiple classes is used to get pose kernel. Finally, decision forests are used to label key pose sequence in real time, and recognition is finished by key pose sequence. The experimental results show that the recognition accuracy of proposed method can achieve at 99.3%, it has better recognition robustness and less recognition time than several other state-of-the-art approaches.