【摘要】 提出了一种基于核技术的求多元区别分析最佳解的K1PMDA算法,并把这一算法应用于人脸识别中。对线性人脸识别中存在两个突出问题:1、在光照、表情、姿态变化较大时,人脸图像分类是复杂的、非线性的;2、小样本问题,即当训练样本数量小于样本特征空间维数时,导致类内散布矩阵奇异。对于前一个问题,可以采用核技术提取人脸图像样本的非线性特征,对于后一个问题,采用加入一个扰动参数的扰动算法。通过对ORL,YaleGroupB以及UMIST三个人脸库的实验表明,该算法是可行的、高效的。更多还原

【Abstract】 A new algorithm, namely kernel machine-based one-parameter multiple discriminant analysis (K1PMDA), to extract optimal discriminant features was proposed and applied to face recognition. There are two problems in linear face recognition: One is that the distribution of face images with different pose, illumination and face expression is complex and nonlinear. The other is the small sample size (S3) problem. This problem occurs when the number of training samples is smaller than the dimensionality of feature vectors, which results in a singular within-class scatter matrix. For the former, kernel technique can be used to extract nonlinear feature, and for the latter, a disturbed parameter was introduced to overcome S3 problem. Three databases, namely ORL, Yale Group B, and UMIST were selected for evaluation. The results are encouraging.更多还原