基于多特征融合和Boosting RBF神经网络的人脸识别  

Face Recognition Based on Feature Fusion and Boosting RBF Neural Network

作　　者：嵇新浩[1] 丁荣涛[1] 

机构地区：[1]浙江商业职业技术学院信息技术系,杭州310053

出　　处：《计算机科学》2008年第6期196-198,共3页Computer Science

摘　　要：提出一种多特征信息融合的人脸识别方法。应用Zernike矩方法和非负矩阵分解法(NMF)分别提取具有旋转不变性的人脸几何特征和人脸子空间投影系数特征,将这两种具有一定互补性的特征串行融合,得到一个分类能力更强的特征。在此基础上,采用RBF神经网络进行人脸识别。为了提高神经网络的分类准确率和泛化能力,采用Boosting方法进行网络集成。实验结果表明,提出的算法利用较少样本数据即可快速地进行人脸识别。A new face recognition method based on feature fusion was proposed. Firstly, face projection coefficient features were extracted by NMF methods. Then these features were combined with the face rotation invariance Zernike moments features so as to get a new feature which had higher discriminating power. With this new feature, RBF neural network is used to class the face. In order to improve the precision of the RBF neural network for face recognition, boosting algorithm is used to build an integration-neural network. The experiment results show that the algorithm can recognize face fleetly using a few samples.

关 键 词：人脸识别 ZEMIKE矩 非负矩阵分解法 BOOSTING方法 RBF神经网络 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术] TP183[自动化与计算机技术—计算机科学与技术]