基于改进LDA算法的人脸识别  

Face Recognition Based on Improved LDA Algorithm

作　　者：齐兴敏[1] 尹朝庆[1] 李智博[1] 

机构地区：[1]武汉理工大学,武汉430063

出　　处：《计算机与数字工程》2007年第8期31-32,37,共3页Computer & Digital Engineering

摘　　要：提出一种基于改进LDA的人脸识别算法,该算法克服传统LDA算法的缺点,重新定义样本类间离散度矩阵和Fisher准则,从而保留住最有辨别力的信息,增强算法的识别率。实验结果证明该算法是可行的,与传统的PCA+LDA算法比较,具有较高的识别率。This article gives a new algorithm of face recognition based on LDA for face recognition.It overcomes the weakness of the traditional LDA algorithm by redefining the matrix between samples and the Fisher rules.Thus the most important information is reserved and the recognition rate is improved.The test result also shows this algorithm is feasible and it has a higher recognition rate compared to traditional way of PCA+LDA.

关 键 词：LDA PCA 识别率 样本类间离散度矩阵 样本类内离散度矩阵 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]