摘要： 针对基于行列投影特征融合的二维线性判别分析中存在的问题，提出一种行列特征复融合的人脸识别算法。通过二维线性判别分析获得行和列的特征矩阵融合成一个复特征矩阵，从复特征矩阵重提取最具分类能力的系数组成特征向量。利用AT&T和AR人脸数据库对该算法进行性能测试，结果表明该算法具有较高的识别率。

关键词: 人脸识别, 二维线性判别分析, 小样本容量问题, 特征融合

Abstract: To avoid the problems which result from twice projects of Two-Dimensional Linear Discriminant Analysis((2D)2LDA), this paper proposes a novel method named complex integration of row and column features based on Two-Dimensional Linear Discriminant Analysis (CI(2D)2LDA). A complex feature matrix is integrated from both row feature matrix and column feature matrix, which are calculated by CI(2D)2LDA. According to their discriminative abilities, a feature vector is extracted from the complex feature matrix. The method’s performance is tested with AT&T and AR face databases. Results show the method is effective and achieves higher recognition rate.

Key words: face recognition, Two-Dimensional Linear Discriminant Analysis((2D)2LDA), small sample size problem, feature integration

中图分类号: 

• TP391.4