摘要： 针对传统线性判别分析中存在的问题，提出一种基于改进零空间法的人脸识别方法，利用奇异向量的稳定性对零空间上的类间散度矩阵投影进行奇异值分解，并对奇异值进行尺度化处理。在ORL和Yale人脸库中对该方法进行性能测试，实验结果表明，该方法是有效的，且具有较高的识别率。

关键词: 线性判别分析, 人脸识别, 小样本问题, 零空间, 奇异值分解

Abstract: Aiming at the problems existed in traditional Linear Discriminant Analysis(LDA), a novel method for face recognition based on improved null space approach is proposed. The projection of within-class scatter vector on the null space is conducted with singular value decomposition by using the stability of singular vector. The singular value is also conducted with measure of treatment. The performance of this method is tested in both ORL and Yale face databases. Experimental results show this method is effective and achieves higher recognition rate.

Key words: Linear Discriminant Analysis(LDA), face recognition, Small Sample Size(SSS) problem, null space, singular value decomposition

中图分类号: 

• TP391