基于动态时间规整算法的虚拟沙画手势识别方法

Method for recognizing gesture of virtual sand painting based on dynamic time warping algorithm

针对目前虚拟沙画手势识别率不高且沙画手势模板较少的情况,提出了1种基于动态时间规整算法的虚拟沙画识别方法,采用Leap Motion获取原始数据,通过沙画手势周期特征值提取方法引入1种新的沙画手势特征值向量组合;进而建立包括12种虚拟沙画手势的模板库;最后通过动态时间规整算法计算测试手势和模板手势的相似度,由相似度差异得出虚拟沙画手势识别的结果。实验表明,所提方法识别虚拟沙画手势准确率高,不同沙画手势间的耦合性较低。

To solve the issues that the virtual sand gesture recognition rate is not high and the number of sand gesture templates is small, a virtual sand painting recognition method based on dynamic time warping algorithm is proposed. First, Leap Motion is used to obtain the original data, and a new sand painting gesture eigenvalue vector combination is introduced according to themethod of extracting gesture period eigenvalue. And l^hen, a template library is created containing 12 kinds of virtual sand paint-ing gestures. Finally, the similarity between the test gesture and the template gesture is calculated through the dynamic time reg-ularization algorithm, and the result of the virtual sand gesture recognition s obtained show that this method can identify the virtual sand gestures with high accuracy and different sand gestures have good low cou-pling.