计算机工程与应用 ›› 2017, Vol. 53 ›› Issue (1): 244-248.DOI: 10.3778/j.issn.1002-8331.1604-0205

• 工程与应用 • 上一篇    下一篇

多点触控的沙画手势识别

席晓晨,况立群,韩  燮,杨晓文   

  1. 中北大学 计算机与控制工程学院,太原 030051
  • 出版日期:2017-01-01 发布日期:2017-01-10

Sand painting gesture recognition based on multi-touch

XI Xiaochen, KUANG Liqun, HAN Xie, YANG Xiaowen   

  1. Computer and Control Engineering, North University, Taiyuan 030051, China
  • Online:2017-01-01 Published:2017-01-10

摘要: 针对多点触控手势间接指令问题,提出了基于多点触控的沙画手势识别系统,该识别系统由时间、空间、形状信息控制。提出一种手势图形建模方法,测量手势的笔划之间的空间和时间关系。采用聚类算法标记手势图形中笔划的形状信息作为局部形状特征;利用基准方法HBF49特征提取全局形状特征。通过一组有10种不同多点触控的沙画手势的数据集评估基于多点触控的沙画手势识别系统,使用图嵌入方法和SVM分类进行手势识别,识别的准确率达到94.75%。实验结果证明,此研究对完成基于多点触控的沙画虚拟系统有重要作用。

关键词: 多点触控, 沙画手势, 手势识别, 图形建模, 图嵌入

Abstract: Aiming at the problems of multi-touch gesture indirect command, a sand painting gesture recognition system based on multi-touch has been presented, which is controlled by temporal, spatial and shape information. Firstly, a gesture graphic modeling method is proposed to measure the spatial and temporal relationship between the gesture strokes. Then, the clustering algorithm is used to mark the shape information of the strokes in the gesture graph as the local shape feature and the benchmarking tool HBF49 is used to extract the global shape feature. Finally, the sand painting gesture recognition system based on multi-touch is evaluated by a collection of data with 10 different multi-touch sand painting gestures, using the graph embedding method and SVM classifier to identify gestures, achieving a recognition rate of 94.75%. The experiments show that the research plays an important role in completing the sand painting virtual system based on multi-touch.

Key words: multi-touch, sand painting gesture, gesture recognition, graph modeling, graph embedding