Abstract: This article first introduces the method used on face authentication with BP neural networks，then proposes a new model（Cs_Pyramid） which is the combination of the Cs_PCA and pyramidal neural network based on Cs_PCA.The traditional BP neural networks are restricted to the dimensions of input samples，so it needs to reduce the dimensions by kinds of methods before training.But，restricted to these methods，lots of information will be lost and it will influence the effect of the face authentication.So this paper brings Cs_Pyramid under this situation.In the same way，Cs_BP is proposed from BP with dimensionality reduction by PCA based on Cs_PCA.Besides，compares the experimental results between Cs_BP and Cs_Pyramid on ORL face database under the LAUSANNE protocol.The results indicate that the Cs_Pyramid is better than Cs_BP.

Key words: pyramidal neural networks, Principal Component Analysis（PCA）, BP neural networks, face authentication, LAUSANNE protocol

摘要： 首先提出BP神经网络在人脸验证上的应用方法，并在Cs_PCA方法的基础之上，提出一种“Cs_PCA+塔式神经网络”的人脸验证新模型（Cs_塔式）。传统的神经网络受到输入样本维数大小的限制，必须经过各种降维处理才能加以训练，受各种降维方法的限制，在降维过程中会丢失相应的数据信息，因此验证效果受到影响。针对此种情况提出了Cs_塔式方法，利用同样的方法，普通BP网在Cs_PCA基础上，利用PCA方法降维构成Cs_BP模型，并且遵照LAUSANNE协议在ORL人脸库上与Cs_塔式模型进行了比较。结果表明，塔式网络有着更好的验证效果。

关键词: 塔式神经网络, 主成分分析, 前向BP网, 人脸验证, LAUSANNE协议