基于独立分量分析和神经网络的人脸识别方法  

Face Recognition Method Based on Independent Component Analysis and Neural Network

作　　者：王明祥[1] 莫玉龙[1] 

机构地区：[1]上海大学通信与信息工程学院,上海200072

出　　处：《计算机工程与应用》2004年第5期4-6,25,共4页Computer Engineering and Applications

基　　金：国家自然科学基金资助项目(编号:69875009)

摘　　要：提出了一种独立分量分析(ICA)和BP神经网络相结合的人脸识别方法(ICABP法),其中人脸图象独立基的提取采用快速独立分量分析方法(FastICA),BP网络采用改进的三层网络结构。该方法将ICA的空间局部特征提取功能和BP神经网络的自适应功能有机地结合起来,增强了系统的鲁棒性。实验表明,ICABP法明显提高了人脸的识别率。A new face recognitio n method combining independent component analysis(ICA)and BP neural network is proposed in this paper.The FastICA algorithm is used to derive the independent sources out of the face images and the improved BP neural network with three layers is used to classify faces.In this paper,the local features extraction o f ICA and the adaptability of BP neural network are combined reasonably.The rob ustness of the system is enhanced greatly.The experiment results show that the face recognition rate of ICABP method is improved obviously.

关 键 词：独立分量分析 主分量分析 神经网络 人检识别 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]