摘要
针对最大间距准则算法中训练样本类内平均值并不能对类内中心做精确估计的问题,提出一种基于中间值的最大间距准则特征提取方法.首先应用样本中间值代替样本的平均值来重新定义类间散度矩阵和类内散度矩阵,然后根据最大间距准则思想得到最优投影矩阵,最后利用三阶近邻分类器进行分类识别.在ORL、Yale和FERET人脸图像库上的仿真实验结果表明,该方法不仅提高了人脸识别率,而且具有较强的鲁棒性.
Aimed at the problem that the intra-class average of training samples in the maximum margin criterion algorithm fails to make accurate estimations of the intra-class center, a new feature extraction method based on the median maximum margin criterion was developed. The sample median was adopted in place of the sample average to redefine inter-class divergence matrix and intra-elass divergence matrix,on the basis of the principle of the maximum margin criterion the optimal projection matrix was obtained,and the third-order neighboring classifier was utilized for classification and identification. Experimental results on face databases ORL, Yale and FERET show that the proposed method improves the human face recognition rate, having strong robustness.
出处
《甘肃科学学报》
2014年第4期21-24,54,共5页
Journal of Gansu Sciences
基金
陕西省教育厅科研计划项目(2013JK0597)
陕西省教育科学"十二五"规划课题(SGH12443)
商洛学院科研基金项目(12SKY010)
关键词
中间值
最大间距准则
人脸识别
特征提取
Median
Maximum margin criterion
Face recognition
Feature extraction
