文中利用Haarlike模板来对不同族群的人脸进行特征提取.利用Haarlike模板生成算法筛选出用于分析人脸特征的Haarlike模板,对构建的人脸数据库进行特征提取,利用C5.0、C&R Tree、BP NN和SVM对训练样本进行学习,并对测试集进行分类和预测,其对维族、藏族、壮族的识别精度分别为80%、74%、88%.为人脸的民族特征识别提供了一种高效快速的方法.
This paper presents a way using Haarlike template to extract face feature on different ethnic groups.Using the Haarlike template generation algorithm searches the templates for the analysis of facial features,completed feature extraction on our multi-ethnic face database,uses C5.0,CR Tree、BP NN and SVM classifier to study the training samples,classifies and predicts the test dataset.The identification accuracy on the ethnic of Uighur,Tibetan,Zhuang is 80%、74%、88% Separately.It also provides a rapid and efficient method for the facial Ethnic feature identification.