【作者】 韩柯； 朱秀昌； 冯荃；

【Author】 HAN Ke ZHU Xiu-chang FENG Quan (College of Telecommunications & Information Engineering,Nanjing University of Posts & Telecommunications,Nanjing 210003)

【机构】 南京邮电大学通信与信息工程学院；

【摘要】 提出了一种新的基于二维 DCT 变换与 Elman 神经网络相结合的方法进行人脸识别。首先,利用二维 DCT 变换将人脸图像由空间域变换到频域,并提取二维 DCT 系数的低频成分作为人脸特征,然后使用人脸训练样本对 Elman 神经网络进行训练来调整网络的权值。对于待识别样本,在利用二维 DCT 变换进行特征提取之后, 采用训练后的 Elman 神经网络作为分类器进行模式分类。该方法在南京理工大学603(NUST603)人脸图像库中进行了实验,实验结果显示了该方法的可行性和有效性。更多还原

【Abstract】 A new method combining two-dimensional Discrete Cosine Transform(DCT)and Elman neural network is proposed for face recognition in this paper.First,two-dimensional DCT is applied to transform face images from the spatial domain to the frequency domain,and low-frequency coefficients of two-dimensional DCT coefficients are extracted as face features.Then,face image training samples are used to train Elman neural network in order to adjust its weight values.For a given test face image sample,the trained Elman neural network is employed to classify the test sample after feature extraction using two-dimensional DCT.The proposed method is evaluated on the Nanjing University of Science and Technology 603(NUST 603)face image database.Experimental results indicate that the method in this paper is feasible and effective.更多还原