摘要
针对最大间距准则在人脸特征提取过程中的不足,提出一种统计不相关的加权最大间距准则人脸特征提取方法。首先对最大间距准则的类间散度矩阵和类内散度矩阵加乘权函数。然后在准则函数中利用双参数调节类间散度和类内散度对特征抽取的影响力。最后通过Schmidt正交化得到统计不相关的最佳鉴别矢量集。在ORL和Yale人脸图像库上的仿真实验结果表明,克服了最大间距准则的缺点,提高了人脸识别率。
According to the shortage of maximum margin criterion in the face feature extraction process, a sta- tistical uncorrelated weighted maximum margin criterion face feature extraction method is presented. Firstly, weight function by class between divergence matrix and class inside divergence matrix is added. And then double parame- ters in the criterion function to control the influence of the feature extraction on class between divergence and class inside divergence are used. Finally statistical uncorrelated optimal identification vector set through the Schmidt orthogonalization got is. Experimental results on ORL face database and Yale face database show that this method overcomes the shortcomings of the maximum margin criterion, and improve the face recognition rate.
出处
《科学技术与工程》
北大核心
2013年第9期2566-2571,共6页
Science Technology and Engineering
基金
陕西省教育厅科研计划项目(11JK0512)
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(11SKY003
12SKY010)
商洛学院教育教学改革项目(10jyjx02006)资助
关键词
人脸识别
加权最大间距准则
统计不相关
特征提取
face recognition weighted maximum margin criterion satistieal uneorrelated feature extraction
