基于优化的LDA算法人脸识别研究

庄哲民; 张阿妞; 李芬兰

doi:10.3724/SP.J.1146.2006.00319

基于优化的LDA算法人脸识别研究

doi: 10.3724/SP.J.1146.2006.00319

基金项目:

广东省自然科学基金项目(032030)资助课题

计量
- 文章访问数: 3884
- HTML全文浏览量: 84
- PDF下载量: 3031
- 被引次数: 0
出版历程
- 收稿日期: 2006-03-20
- 修回日期: 2006-08-21
- 刊出日期: 2007-09-19

Based on an Optimized LDA Algorithm for Face Recognition

摘要

摘要: 提取低维人脸特征是人脸识别系统中极其关键的一步。线性判别分析(LDA)是一种较为普遍的用于特征提取的线性分类方法。本文提出了一种优化的LDA算法，该方法克服了传统的LDA算法用于人脸识别时存在的问题：通过重新定义样本类间离散度矩阵使传统的Fisher准则能够最优化，克服了边缘类对选择最佳投影方向的影响；同时，利用因数分解的方法避免了对矩阵求逆，解决了小样本问题。依据经验选取适当的e值，得到最佳的识别效果。实验结果表明，人脸识别效果优于传统LDA方法。
- 线性判别分析(LDA);人脸识别;类间离散度;类内离散度;特征提取
Abstract: Extracting the most discriminant low-dimensional face feature is an extremely critical step in Face Recognition (FR) systems. Linear Discriminant Analysis (LDA) is one of the most popular linear classification techniques for feature extraction. An optimized LDA algorithm is introduced to overcome questions existing in the traditional LDA algorithm for FR in this paper. The between-class scatter matrix is redefined in order to make the traditional Fisher criterion optimal and eliminate the effect that the edge of class has on selecting the optimal projection; at the same time, it avoids computing the inverse of matrix by means of factorization, and solves the Small Sample Size (SSS) problem. Adopting experiential method, the appropriate value of e is found, and then the optimal effect of face recognition is got. Experimental results show the recognition rate of this method is superior to the traditional LDA.

HTML全文

参考文献(1)

Chellappa R, Wilson C, and Sirohey S. Human and machine recognition of faces: A survey[J].Proc. IEEE.1995, 83(5):705-741[2]Tplba A S, El-Baz A H, and El-Harby A A. Face recognition: A literature review. J. of Signal Processing, 2005, 2(1): 88-103.[3]Turk M and Pentland A. Eigenfaces for recognition[J].J. of Cognitive Neuroscience.1991, 3(1):71-86[4]Belhumeur P N, Hespanha J P, and Kriegman D J. Eigenfaces vs. fisherfaces: Recognition using class specific linear projection. IEEE Trans on Pattern Analysis and Machine Intelligence, 1997, 19(7): 711-720.[5]Chen L F, Liao M, and Lin J C, et al.. A new LDA-based face recognition system which can solve the small samples size problem[J].J. of Pattern Recognition.2000, 33(10):1713-1726[6]Yu H and Yang J. Direct LDA algorithm for high dimensional data with application to face recognition[J].J. of Pattern Recognition.2001, 34(10):2067-2070[7]Huang R, Liu Q S, and Lu H Q, et al.. Solving the small sample size problem of LDA. IEEE Proceedings of the 16th International Conference on Pattern Recognition, Canada, Quebec 2002, 3: 29-32.[8]Lotlikar R and Kothari R. Fractional-step dimensionality reduction[J].IEEE Trans. on Pattern Analysis and Machine Intelligence.2000, 22(6):623-627[9]Loog M, Duin R P W, and Haeb-Umbach R. Multiclass linear dimension reduction by weighted pairwise fisher criteria[J].IEEE Trans. on Pattern Analysis and Machine Intelligence.2001, 23(7):762-766[10]Martinez A M and Zhu M. Where are linear feature extraction methods applicable? IEEE Trans[J].on Pattern Analysis and Machine Intelligence.2005, 27(1):1934-1944

施引文献

资源附件(0)

访问统计