A Discriminant Model for the Pattern Recognition of Linearly Independent Samples

作者：Wu, Wei; Ahmad, M. Omair^*

来源：Circuits, Systems, and Signal Processing, 2012, 31(2): 669-687.

DOI：10.1007/s00034-011-9337-2

• Pattern recognition
• Face recognition
• Linear discriminant analysis
• Discriminant model
• Linearly independent samples
• Small sample size problem

Fisher's linear discriminant analysis (FLDA) has been used in many pattern recognition applications. However, this method cannot be applied for solving the pattern recognition problems when the within-class scatter matrix is singular, i.e., when the so-called small sample size problem occurs. Many FLDA variants that have been proposed in the past to circumvent this problem either suffer from excessive computational load when patterns have a large dimension or lose some useful discriminant information. In this paper, a new systematic framework for the pattern recognition of datasets with linearly independent samples is developed. Within this framework, a discriminant model, in which the samples of the individual classes of a dataset lie on parallel hyperplanes and project to single distinct points of a discriminant subspace of the underlying input space, is shown to exist. Based on this model, three algorithms that do not encounter the adverse effects of the small sample size (SSS) problem are developed to obtain such a discriminant subspace for a given dataset with linearly independent samples. A kernelized algorithm is also developed for the discriminant analysis of datasets for which the samples are not linearly independent. Simulation results are provided to examine the validity of the proposed discriminant model and to demonstrate the effectiveness, both in terms of complexity and classification accuracy, of the linear and nonlinear algorithms designed based on the proposed model.

• 出版日期2012-4
• 单位深圳清华大学研究院