ECAI 2004 Conference Paper

[full paper]

Face Recognition Using Novel LDA-Based Algorithms

Guang Dai, Yuntao Qian

Facial feature extraction with enhanced discriminatory power plays an important role in face recognition (FR) applications. The linear discriminant analysis (LDA) is a powerful tool used for dimensionality reduction and feature extraction in FR tasks. However, the classification performance of traditional LDA is often degraded, due to two factors: 1) their classification accuracies suffer from the small sample size problem (SSSP), which widely exists in FR; 2) their Fisher discriminant criterions are not directly related to the classification ability. Recently, called direct fractional-step LDA (DF-LDA) algorithm has been proposed to solve this problem. In this paper, the limitations of DF-LDA are discussed and a novel DFLDA has been proposed to solve those problems. The novel DF-LDA has been tested, in terms of classification accuracy, on the ORL and the UMIST face databases. Results reveal that the proposed method outperforms the previous existing methods, including: the Eigenfaces, Fisherfaces, D-LDA, previous DF-LDA, and EFM methods.

Keywords: Keywords: Face recognition (FR), feature extraction, linear discriminant analysis (LDA), small sample size, problem (SSSP).

Citation: Guang Dai, Yuntao Qian: Face Recognition Using Novel LDA-Based Algorithms. In R.López de Mántaras and L.Saitta (eds.): ECAI2004, Proceedings of the 16th European Conference on Artificial Intelligence, IOS Press, Amsterdam, 2004, pp.455-459.