On a Face Recognition by the Modified Nonsingular Discriminant Analysis for a Ubiquitous Computing

This paper presents an efficient face recognition by the modified nonsingular discriminant analysis for a ubiquitous computing. It is popular to extract discriminant features using Fisher linear discriminant analysis (LDA) for general face recognition. In this paper, we propose the modified nonsingular discriminant analysis in order to overcome the problem of small sample size and prone to be unrealizable due to the singularity of scatter matrices. The scatter matrix of transformed features is nonsingular. From the experiments on facial databases, we find that the modified nonsingular discriminant feature extraction achieves significant face recognition performance compared to other LDA-related methods for a limited range of sample sizes and class numbers. Also, recognition by the modified nonsingular discriminant analysis by using TMS320C6711 DSP Vision Board is set to highlight the advantages of our algorithm.