基于关键帧和姿态描述语言的实时手势识别方案.pdf

第 38卷 第 3 期 湘 潭 大 学 自 然 科 学 学 报 Vol. 3 8 N 〇. 3 2016 年 9 月 Natural Science Journal of Xiangtan University Sep. 2016 基于关键帧和姿态描述语言的实时手势识别方案 包莹莹〃 ， 王华君、 徐平平 2 (1.无锡太湖学院工学院，江 苏 无 锡 214064) (2.东南大学信息科学与工程学院，江 苏 南 京 211189) [摘要 ] 针对现有人体动作识别方法的准确性和实时性不强等问题，提出一种基于关键帧和姿态描述语 言（PDL) 的实时手势识别方案 . 首先，利 用 Kinect摄像机获取手势视频帧，并 通 过 N I T E 工具获得骨架关 节点的三维坐标 . 然后，从视频序列中找出最能代表手势类型的关键帧 . 最后，提出一种姿态描述语言， 根据关键帧中关节点的位置和角度信息，通过一系列规则对手势进行识别 . 实验结果表明，该方法具有较 高的识别率和良好的实时性 . 关键 词 ：手势识别；骨架关节点；关键帧;姿态描述语言 中图分类号：TP391 文献标识码：A 文章编号 = 1000 -5900(2016)03 - 0078 - 06 Real-time Gesture Recognition Scheme Key Frame and Posture Description Language BAO Ying-ying1* , WANG Hua-jun1 , XU Ping-ping2 (1. School of Engineering, Taihu University of Wuxi, Wuxi 214064; 2. School of Information. Science and Engineering, Southeast University, Nanjing 211189 China) 【Abstract】 For the issue that the existing human motion recognition methods are poor in accuracy and real­ time performance, a real time gesture recognition scheme based on key frame and posture description lan­ guage (PDL) is proposed. Firstly, Kinect camera is used to obtain the gesture video frame, and the 3D co­ ordinates of the frame joint point are obtained by the NITE tool. Then, the representative keyframes for gesture type are found out from the video sequence. Finally, a kind of posture description language is pro­ posed, which uses a series rules to recognize gesture according the position and angle of the joint points in the key frame. Experimental results show that the proposed method has higher recognition rate and good re­ al-time performance. Key words ：human action re