基于小波变换和独立成分分析的人脸识别  

Face recognition based on wavelet transform and independent component analysis

作　　者：刘嵩[1,2] 罗敏[2] 向军[2] 张国平[1] 

机构地区：[1]华中师范大学物理科学与技术学院,武汉430079 [2]湖北民族学院信息工程学院,湖北恩施445000

出　　处：《华中师范大学学报（自然科学版）》2012年第2期166-169,共4页Journal of Central China Normal University：Natural Sciences

基　　金：国家自然科学基金项目(61040006);恩施州科技局项目

摘　　要：针对传统的独立成分分析算法对光照、表情、姿态等敏感的不足,提出了一种结合小波变换和独立成分分析的人脸识别方法.人脸图像首先经过小波变换后选取低频子图像进行独立成分分析,提取人脸图像特征,最后根据最近邻分类器分类.分析了样本数目、小波分解级数对平均识别率和识别时间的影响.基于ORL人脸数据库的实验结果证明了本方法在识别性能方面相对于单一方法的优越性.A method of face recognition based on wavelet transform and independent component analysis is proposed to decrease the influence of illumination, facial expressions, posture and other factors over the recognition rate. Firstly, wavelet decomposition is used as a pre-processing method, and the low-frequency face image is choised as a sub-image, then the feature of sub-image is extracted by ICA. Finally, the nearest neighbor classifier is used to recognize different faces from the ORL face database. The effect on the samples number and wavelet decomposition levels is analyzed in the aspects of recognition time and recognition rate. Experimental results show that the proposed method improved the recognition performance in comparison with ICA.

关 键 词：人脸识别 独立成分分析 小波变换 最近邻分类器 特征提取 非高斯性 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]