中文摘要：

目的：提出一种利用多进制小波分析提取人脸特征的识别方法。方法：对人脸图像进行八进制Haar小波分解后提取低频信息与高频信息,组成人脸特征库,然后利用欧式距离匹配对不同人脸进行分类识别。结果：经Yale人脸数据库中90幅人脸图像的实验验证,总识别率达到了96%。结论：该方法能有效地消除因人脸图像的表情变化和少许遮掩带来的识别误差,为人脸精准识别提供基础。

英文摘要：

Objective To propose a face features extraction method using multi-band wavelet analysis. Methods Haar 8- band wavelet was used to decompose the face image, and low- and high-frequency information was extracted to form the face feature library, then euclidean distance matching was used for classified face recognition. Results Trials for 90 images in Yale face library proved that the recognition rate reached 96%. Conclusion The method can be used for precise face recog- nition, with the errors by expression variation and concealing eliminated.