摘要
为解决设备交互不便、光照变化对识别效果影响较大等问题,提出一种基于深度图像的手势识别方法。利用Ostu分割法将手部区域分割出来并对手形进行提取,得到轮廓、掌心、指尖等手部形态特征。在此基础上,提取手势特征,如手指间角度、是否存在凸包点、手指是否弯曲等,检测并实时识别10种数字手势。实验结果表明,该方法可在复杂背景、肤色干扰等变化条件下快速准确识别各种数字手势,识别率较高,具有良好的实时性和鲁棒性。
To solve the problem that the device interaction is inconvenient and the illumination changes have a great influence on the recognition effects,a gesture recognition method based on depth information was proposed.The Ostu segmentation method was used to segment the hand region and extract the hand shape to obtain the hand shape features such as contour,palm and fingertip.On this basis,the gesture features were extracted,such that what were the angles between the fingers,whether there was a convex point,whether the fingers were bent,e.g.,last 10 kinds of gestures were detected and recognized in real time.Experimental results show that the proposed method can quickly and accurately identify various digital gestures under the conditions of complex background and skin color interference,and it has high recognition accuracy,good real-time performance and high robustness.
作者
时梦丽
张备伟
刘光徽
SHI Meng-li;ZHANG Bei-wei;LIU Guang-hui(School of Information Engineering,Nanjing University of Finance and Economics,Nanjing 210023,China)
出处
《计算机工程与设计》
北大核心
2020年第7期2057-2062,共6页
Computer Engineering and Design
基金
国家自然科学基金项目(60802087)
江苏省研究生科研与实践创新计划基金项目(KYCX18_1392)。
关键词
体感设备
深度图像
阈值分割
手势特征
实时识别
Kinect
depth image
threshold segmentation
gesture feature
real-time recognition