摘要
提出了基于手势伸展方向以及手势图像高阶NMI值的特征向量进行识别分类的算法。首先,对采集到的手势图像进行预处理与二值化,然后按手势的伸展方向进行粗分类,之后,对每幅图像提取图像的1阶和4阶NMI值,这样将一幅图像表示成一个有17维分量的特征向量。识别时用k近邻原则进行判别,识别率达到86.6%。
An alphabet gesture recognition algorithm based on orientation of gesture and high-order NMI feature of the image is presented. Firstly, the images are pre-processed to get the binary image. Then the system grossly classifies the alphabet gestures according to the gesture orientation and proceeds to calculate the high-order NMI feature of the gesture image. Lastly it recognizes the unknown samples by orientation and NMI feature. Experiments show that this method is efficacious and easy to operate.
出处
《计算机工程》
EI
CAS
CSCD
北大核心
2005年第14期164-166,共3页
Computer Engineering
基金
上海市自然科学基金资助项目(02ZD14053)