摘要
本文提出了一种基于形状特征的静态手势识别算法.对分割出来的手势二值图像进行细化处理,根据细化骨骼恢复出手指形状的候选区域,提取形状特征并进行手势识别,即判断手指个数从而对手势数字0-5进行识别.本文在Matlab2012平台下对采集的中型数据库样本进行识别,平均识别率达到97.91%,平均识别速度为2.14s.同时,对实时采集的数据结果进行测试,实验结果具有一致性.
This paper presents a static hand gesture recognition algorithm based on shape features. The proposed algorithm segments the binary images of hand gesture, adopts thinning procedure to extract the hand skeleton, reconstructs the candidate fingers and extracts the shape features. Finally, the features are classified into 6 categories, namely 0 to 5. The experimental result on a medium sized database shows that the average recognition rate is 97.91% and the average recognition time is 2. 14s on the Matlab 2012 platform. Real time tests were also conducted and the result is the same.
出处
《北方工业大学学报》
2015年第3期43-48,24,共7页
Journal of North China University of Technology
基金
北京市教委面上项目(KM201510009005)
关键词
手势识别
肤色检测
手指重建
形状特征
gesture recognition
skin detection
finger restoration
shape feature
