基于空间支持向量域分类器的人脸识别

摘要

A new approach for face recognition based on wavelet, singular value decomposition and Space Support Vector Domain Classifier(SSVDC) is presented. SSVDC is applied to classify singular value feature vector of face image, and the distance from the test sample to each hyper sphere center is computed. The attribution is judged according to distance from the test sample to hyper sphere central. Experiment result on (Olivetti Research Laboratory,ORL) database demonstrates that recognition accuracy is 97.5％.%提出了一种基于小波变换、奇异值分解与空间支持向量域分类器相结合的人脸识别方法.在使用空间支持向量分类器对不同人脸图像的奇异特征向量进行分类时,计算所测样本到各个超球球心的距离,并根据其与超球半径的关系来判断其所归属.并在ORL人脸数据库中进行实验.实验表明提出的人脸识别方法识别精度可达97.5%.