摘要
该文提出了一种将支持向量机(SVM)与神经网络相结合的方法,得到一种新型 的级联型组合分类器．此组合分类器先利用神经网络或SVM对人脸图像进行预分类,得到不 同性别的两类人脸图像;然后分别针对其中某一类人脸图像进行K-L变换以提取有效特征, 再使用SVM进行细分,得到最终的识别结果．应用该组合分类器方法在本文整合得到的人脸 样本库上进行测试,结果显示该方法不但可以有效地提高识别速度,而且还可以在一定程度上 提高识别率,因此方法是有成效和有价值的．
This paper presents a method of combining the classifiers of SVM with BP network to get a new serial combined classifier which differentiates face images adapting to different genders. Through K-L transform, we distill the effective features for every class of face images and then use SVM to classify these images too and get the last result. Test results on the face images database conformed in the paper show that the method not only can increase the recognition speed, but also can get a higher recognition rate. So this method is effective and valuable.
出处
《嘉兴学院学报》
2005年第6期64-67,共4页
Journal of Jiaxing University
