摘要
人的手势是人们日常生活中最广泛使用的一种交流方式。由于在人机交互界面和虚拟现实环境中的应用,手势识别的研究受到了越来越广泛的关注。但是目前基于单目视觉的手势识别技术中,手势分割要求背景简单或者要求识别者戴着笨重的数据手套。而该文结合了运动信息和基于KL变换的肤色模型,在复杂背景下进行手势分割,与传统的基于RGB肤色模型的手势分割相比,在复杂背景环境下得到了很好的分割效果。在对分割的手势区域进行预处理后,该文使用了一种归一化的傅立叶描述子进行手势的特征提取,相比传统的傅立叶描述子更加准确,最后采用了传统的三层BP网络作为模式识别器,手势训练集和测试集的识别率分别达到了95.9%和95%。
Hand gesture is one of the most popular communication methods in everyday life. Hand gesture recognition research has gained a lot of attentions because of its applications for interactive human - machine interface and virtual environments. But currently, in the vision - based hand gesture recognition, almost all the technologies on hand gesture segmentation are based on simple background or on gloves in special colors. However, this paper presents a method that segments the hand gestures with complex backgrounds through the combination of motion and skin color based on KL Transformation,in contrast with traditional hand gesture segmentation based on RGB color model in some environments. After the pretreatment to hand gesture region, we use a normalized Fourier descriptor, which is more accurate than the traditional Fourier descriptor,to select the hand gesture features and use the traditional 3 levels BP network to perform hand gesture recognition. Finally, the average recognition rate is 95.9% on the training set and 95% on the testing set.
出处
《计算机仿真》
CSCD
2005年第12期158-161,共4页
Computer Simulation
关键词
肤色模型
手势分割
手势识别
归一化的傅立叶描述子
Skin color model
Hand gesture segmentation
Hand gesture recognition
Normalized Fourier descriptor
