摘要
手势识别中,除手的分割是个难题外,在视频流中,手势有效动作起止帧的判定也是一个急待解决的问题。将图像的肤色与运动信息结合,分割出手,建立手势运动的历史图(Motion History Image,MHI),提出了一种MHI与改进的运动能量图(Improved Motion Energy Image,IMEI)结合的机制,判定有效手势运动的起止帧,最后建立有效手势运动的MHI;同时提出改进图像梯度向量算法用于手势运动方向(上,下,左,右)识别。实验表明,对于有效手势运动起止帧的正确判断率一般可达90%以上,有效手势识别率达96%以上,手势交互自然顺畅。
In gesture recognition,hand segment and the first frame of hand motion are two problems.In this paper,skin image combines with hand motion information in order to segment the hand region.Then,construct the Motion History Image(MHI) and Improved Motion Energy Image(IMEI) to determine the first and last frames of effect hand motion.Experiment shows that the ratio is up to 90%.This paper gives an improved image gradient vector(IIGV) to recognize gestures.IIGV gives the recognition ratio is up to 96% and gesture interaction is natural and smooth.
出处
《西南科技大学学报》
CAS
2012年第1期44-47,共4页
Journal of Southwest University of Science and Technology
关键词
肤色概率图
运动历史图
运动能量图
图像梯度向量算法
Color probability map
Motion history image
Motion energy image
Image gradient vector