摘要
本文提出了一种全新的手指书写虚拟文字识别系统,该系统利用摄像头捕捉人手指的运动轨迹,通过手指跟踪及检测算法,恢复出手指虚拟“书写”文字的二维图像数据(称之为虚拟文字),再进行识别输出.文中对单层函数连接神经网络进行改进并用来进行高准确度的运动手指尖的预测及跟踪,给出了一种基于数学形态学及模板匹配的简单有效的手指检测方法,提出了基于运动手指跟踪的“虚拟文字”恢复方法.实验结果表明:系统方案可行,手指跟踪及检测准确率高,对手指“书写”的数字、英文等虚拟文字识别率能达到95%以上.
A virtual character recognition system based on finger tracking is proposed. By capturing the finger movement with a digital camera and tracking the finger-tip position automatically, the 2-D character image （virtual character trajectory） ＂written＂ by finger-tip is reconstructed and recognized. A modified flat functional link neural network is used to predict and track the moving position of the finger-tip. A mathematical morphology and template matching based finger-tip detection approach is pro- posed, and an algorithm to reconstruct and recognize the virtual character is given. Experiments show that the proposed system works very well with high finger tracking and detection accuracy, and achieves more than 95 % accuracy rate for virtual digital and English letters recognition.
出处
《电子学报》
EI
CAS
CSCD
北大核心
2007年第3期396-401,共6页
Acta Electronica Sinica
基金
国家自然科学基金(No.60275005)
Motorola国际合作研究基金(No.D84110)
教育部新世纪优秀人才支持计划(No.NCET-05-0736)
关键词
手指书写
虚拟文字识别
字符识别
人机交互
手指跟踪
finger writing
virtual character recognition
character recognition
human machine interaction
fingertip tracking
