基于稀疏RAM的逼近型神经网络与统计模式识别的人脸识别  

Face Recognition Based on The Approximate Neural Network with Sparse RAM and Statistical Pattern Recognition

作　　者：周兆捷[1] 孙玉霞[1] 吴乐南[1] 

机构地区：[1]东南大学无线电工程系,南京210096

出　　处：《信号处理》2003年第6期517-521,共5页Journal of Signal Processing

摘　　要：本文提出了一种基于稀疏RAM的逼近型神经网络(SN-tuple)与统计模式识别相结合的人脸识别方法,采用首先直接将原始图像数据输入稀疏RAM的逼近型神经网络中进行粗分类,再由统计模式识别方法中的PCA、LDA来进行最终细分类的方法,通过大量的实验证明了该方法的有效性。In this paper a face recognition approach based on the approximate neural network with sparse RAM and statistical pattern recognition is presented. Firstly, the data of original face images are imported into the approximate neural network with sparse RAM to classify these faces roughly. Then the PCA or LDA of statistical pattern recognition is used to classify these faces precisely. Experimental results demonstrate the feasibility of the approach.

关 键 词：稀疏RAM 逼近型神经网络 统计模式识别 人脸识别 稀疏分布存储器 主元分析 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]