期刊论文

中文

计算机工程与设计

1000-7024

2018

39

6

1721-1726

10.16208/j.issn1000-7024.2018.06.039

“十二五”国家科技支撑计划基金项目（2015BAK33B02） 国家自然科学基金项目（61501199、61605054） 华中师范大学中央高校基本科研业务费基金项目（230-20205180014-660） 华中师范大学中央高校基本科研业务费教育科学专项基金项目

本校为第一机构

针对手势识别的准确性受到复杂背景、光照变化、肤色干扰、旋转、平移、缩放等影响的问题,提出一种基于Kinect的实时手势识别方法。使用Kinect体感设备,获取人体的骨骼信息和深度信息,分析人体手部特征,根据人体手部轮廓提取指尖位置特征,结合骨骼关节点位置分析人体手部以及得到的手指特征,检测并实时识别10种计数手势。实验结果表明,该方法能够在复杂背景、肤色干扰以及光照发生变化的条件下获得正确的识别结果,满足手势的旋转、平移和缩放不变性,识别率高,具有良好的实时性和鲁棒性。

To solve the problem of hand gesture recognition precision compromise caused by complex backgrounds,lighting conditions,skin color,rotation,translation and scaling and so on,a robust and efficient real-time gesture recognition method using Kinect was proposed.Depth data and joint information of human body were obtained using a Kinect Xbox device.The features of gestures were computed and the coordinates of fingertips were tracked according to the contour of the hand gesture.In combination with the position of the skeletal joints,the hand and fingertip characteristics were calculated and ten ge...