基于SVD压缩降秩与KDA的人脸识别新方法

摘要

A new method of face recognition based on singular value reducing dimension and kernel discriminant analysis (KDA) features extraction combined with counter propagation network (CPN) neural network is presented in this paper. In this method, it adopts a rule of singular value reduce dimension to get excellent singular value,then take the extracted eigenvalue to make KDA reforming, draw optimal eigenvalue to move forward a single step. Finally, taking the optimal eigenvalue as the neural network' s input value, make use of CPN artificial neural network to recognize face images. This method has a high recognition rate and speed.%文中提出了一种基于奇异值压缩降秩与核判别分析(KDA)变换方法的人脸特征提取新方法,同时结合对偶传播人工神经网络(CPN)对不同的人脸图像进行识别分类.该方法首先采用奇异值分解压缩降秩准则对人脸图像进行择优奇异值的选取,然后对提取后的择优特征值进行核判别分析(KDA)变换,进一步提取人脸图像最优特征值,最后将得到的人脸图像最优特征值作为网络的输入值,利用对偶传播人工神经网络(CPN)对人脸图像进行识别分类.实验结果表明该方法具有较高的识别率和较快的识别速度.