摘要： 利用最大间距准则(MMC)寻求一组最佳鉴别矢量，使投影变化后的特征空间的类间散度最大，类内散度最小，并克服小样本问题。与原MMC相比，新特征提取方法通过对原来的散度加乘权重及对参数的调整，能够在特征提取的同时更好地保持人脸图像的局部流形结构。在ORL人脸库上的实验结果表明，该方法能够更为有效地识别人脸图像，提高识别率。

关键词: 最大间距准则, 局部保持, 特征提取, 人脸识别

Abstract: The purpose of the Maximum Margin Criterion(MMC) is to maximize the inter-class scatter while simultaneously minimizing the intra-class scatter after the projection, and does not suffer from the small sample size problem. Compared with original MMC method, by multiplying the defined weights and regulating the parameter, the new method can still better manifold local structure information. Experimental results on Olivetti Research Laboratory(ORL) face database show that the method can recognize the face images efficiently and enhance the recognition rate.

Key words: Maximum Margin Criterion(MMC), locality preserving, feature extraction, face recognition

中图分类号: 

• TP391