基于单目视觉的实时手语识别系统设计与实现  

Design and Implementation of Real Time Gesture Recognition System Based on Monocular Vision

作　　者：王丽光 张根源[2] 刘子龙[2] 

机构地区：[1]山西省交通厅定额站,山西太原030006 [2]上海理工大学光电信息与计算机工程学院,上海200093

出　　处：《电子科技》2017年第3期130-133,137,共5页Electronic Science and Technology

摘　　要：针对目前单目手语识别系统存在检测效率低、识别速度慢的问题。文中提出了基于YCr Cb空间的肤色模型与背景差分模型融合的高效检测算法,从而提高了手势的检测效率;且为了提高识别手势的速度,采用交叉相关算法计算特征相关度识别手势。通过进行8种常用手势测试模型与标准模板库的对比实验,结果验证了本系统具有较高的识别速度,且识别率达到85%以上。In recent years, as the rapid development of computer application and the popularization of the com- puter vision theory, human - computer interaction technology has become as a hot mainstreamtechnology and comput- er science . In order to establish a fast and high efficient gesture language recognition system, this paper has proposed an efficient detected algorithmwhich is composed by the YCrCb skin - color - detect model and the background differ- ence model. And then the gesture pattern recognition is achieved based on the cross correlation algorithm, which is improve the recognition speed. Finally the experiments verify the real - time of the system, and recognition rate reached more than 85% which means thesystem is practical.

关 键 词：手语识别 人机交互 运动轨迹 模式识别 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]