Isomap与径向基神经网络在人脸识别系统中的应用

Application of Isomap and RBF Neural Network in Face Recognition System

基于非线性降维算法的容特征映射与径向基神经网络的快速性,提出了基于Isom ap与径向基(RBF)神经网络的图像识别方法,降维方法用测地距离取代传统的欧式距离,有助于挖掘高维数据的内在结构,径向基神经网络能够快速模拟对象数据集,识别真假图像。同时该方法结合了频谱分析对初始图像进行预处理,减少了计算量。实验结果表明该方法能快速识别真假图像,提高识别率。

Real and fake face recognition is important to face recognition system. Based on isometric feature mapping （Isomap） and quickness of RBF neural network, a new method is used to analyze large-scale real and fake face data. The cluster algorithm improves the distance measurement between face samples by replacing Euclidean distance with the geodesic distance. RBF neural network provides high speed to simulate samples and give the right outcome. The method preprocesses the data based on frequency analysis and provides less computation. The experiment shows that this algorithm provides the fast recognition of real and fake face and improves recognition rate.