摘要
提出了一种新的人脸特征提取方法,该方法采用DCT对人脸图像进行降维和去噪,并通过KDA提取人脸特征。基于该特征,采用NN分类器,对ORL人脸库进行分类识别,仅用28个特征平均识别率就达到97.3%,“留一法”识别率为99.5%。仿真结果表明:该方法有效地滤除了人脸图像中的高频干扰信息,明显增强了特征的辨别能力,同时显著地降低了特征维数和计算复杂度。
A novel face feature extraction method is presented in this paper. In this method, the raw face images are denoised by DCT, and dimension reduced features are obtained, then the KDA is performed on the feature vectors to enhance discriminant power. Finally, the NN classifier is selected to perform face classification. The experimental results on ORL face database show that the proposed method achieves an average recognition accuracy of 97.3% using only 28 features and the ＇leave one out＇ recognition rate is 99.5%. Moreover, the dicriminant power is enhanced effectively, and the computing complexity and feature dimensions are reduced greatly.
出处
《电子科技大学学报》
EI
CAS
CSCD
北大核心
2006年第4期450-453,共4页
Journal of University of Electronic Science and Technology of China
基金
江苏省高校重点实验室基金资助项目(KJS03036)
关键词
人脸识别
核辨别分析
最近邻分类器
face recognition
kernel discriminant analysis
nearest neighbor classifier
