应用几何矩和边缘检测的手势识别算法  

Gesture Recognition Algorithm Based on Invariant Moment and Edge Detection

作　　者：何阳清[1] 葛元[1] 王林泉[1] 

机构地区：[1]上海海事大学计算机系,上海200135

出　　处：《计算机工程》2005年第15期165-166,174,共3页Computer Engineering

基　　金：上海市自然科学基金资助项目(02ZD14053)

摘　　要：提出了一种结合几何矩和边缘检测的手势识别算法。手势图像经过二值化处理后,提取手势图像的几何矩特征,取出几何矩特征7个特征分量中的4个分量,形成手势的几何矩特征向量。在灰度图基础上直接检测图像的边缘,利用直方图表示图像的边界方向特征。最后,通过设定两个特征的权重来计算图像间的距离,对30个字母手势进行识别,识别率为90%。A gesture recognition algorithm based on two features of invariant moment and edge detection is presented. After preprocessing, 4 of 7 invariant moments are extracted on binary image. By edge detection, histogram is formed to describe the edge information. Finally, the recognition of 30 letter gestures is performed by computing distance, in which different coefficients are set to these two features. The recognition rate of 90% is achieved.

关 键 词：手势识别 几何矩 边缘检测 边界方向直方图 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]