【摘要】 人脸具有丰富的表情变化,而且受光照强度、成像角度和成像时间等诸多因素的影响,这些因素都给人脸自动识别造成很大的困难。针对这些问题,笔者提出了一种局部线性鉴别分析(LLDA:Locally L inearD iscrim inantAnalysis)的非线性鉴别分析方法,其根本思想是:全局非线性数据结构可由局部线性和局部结构的线性组合表示。样本的特征矢量通过线性转换构成局部特征子空间,使类间散度最大而类内散度最小。该方法适用于多类非线性鉴别。实验表明,在低维子空间、姿态变化和单视图表示的人脸识别中是很有效的。更多还原

【Abstract】 Face recognition is affected by human faces rich expression,the intensity of ligtht,the angle and the time of face image created,all of there make the recogniton difficult.To deal with these problems,we present a method of nonlinear discriminant analysis called "Locally Linear Discriminant Analysis".The underlying idea is that global nonlinear data structures are locally linear and linearly aligned.Feature vectors are projected into each local feature space by linear transformations,which are found to yield locally linearly transformed classes that maximize the between-class covariance while minimizing the within-class covariance.Our method is for multiclass nonlinear discrimination.The transformation functions facilitate robust face recognition in a low-dimensional subspace,under pose variations,using a single model image.更多还原