摘要
一般的人脸检测在运行时间及检测率上都不能得到很好的保证.本文提出了基于离散余弦变换的支持向量机的人脸检测方法,利用离散余弦变化后的系数作为支持向量机的输入特征,实验表明该方法具有更好的检测效果.实验还表明,在采用离散余弦变换系数作为检测特征值时,检测准确率并不是随着所选取特征值个数的增加而提高.
Usually the runtime and detection rate of face detection were unsatisfactory. In this paper,we presented a method based on discrete cosine transform and Support Vector Machine. It used the coefficient of discrete cosine transformation as the eigenvalue of support vector machine. The experiment shows that it had a better result of detecting faces,and when discrete cosine transform coeffi- cients were userd as eigenvalue,the detection rate was not increased with increasing the number of selected eigenvalue.
出处
《厦门大学学报（自然科学版）》
CAS
CSCD
北大核心
2007年第6期788-791,共4页
Journal of Xiamen University：Natural Science
基金
厦门大学985二期工程项目资助
关键词
人脸检测
离散余弦
支持向量机
face detection
discrete cosine
Support Vector Machine
