摘要
针对人脸表情变化对人脸识别的影响,提出一种结合小波变换(DWT)、特征脸方法(PCA)和线性判别法(LDA)的人脸特征提取新方法。首先将人脸图像通过二维小波变换(2DWT)提取其低频分量,然后将低频图像经过PCA变换映射到一个低维空间,最后在低维空间中利用LDA方法进行人脸特征的提取。通过此方法,采用ORL人脸库和Yale人脸库进行测试,我们可实现更准确的特征提取,并有效解决表情变化对人脸识别的影响问题。实验结果显示,本文方法在提高人脸识别率的同时,也提高了人脸识别速度。
A novel method of Face Feature extraction is presented for the impact of the Face Recognition by Facial Expression changing,which combines Discrete Wavelet Transform （DWT）with newer Principal Components Analysis（PCA） and Linear Discriminant Analysis（LDA）. A face image was first extracted into the low -frequency components image using two -dimensional Discrete Wavelet Transform （2DWT）, then, with the PCA was used to map the low -frequency components image into a low -dimensional feature space, and finally ,with the LDA was used to extract the Face Feature in the low - dimensional feature space. In this way ,using ORL face database and Yale face database to test, more accurate feature was extracted, and the problem of the impact of the Face Recognition effectively solved which had impacted by Facial Expression changing. Experimental results in the Face Feature extraction and Face Recognition demonstrated satisfactory improvement of the recognition rate and recognition speed.
出处
《微计算机应用》
2011年第11期14-19,共6页
Microcomputer Applications
基金
国家自然科学基金(61179011)
福建省自然科学基金资助项目(2010J01327)
关键词
特征提取
人脸表情识别
小波变换(DWT)
特征脸方法(PCA)
线性判别法(LDA)
feature extraction, Face Expression recognition, Discrete Wavelet Transform （DWT）, Principal Components Analysis （PCA） , Linear Discriminant Analysis （LDA）
