By normalizing to improve generalized Foley-Sammon transform in high-dimensional spaces - with application to face recognition

Linear discriminant analysis (LDA) is an effective feature extraction technique for classification. A new LDA-based algorithm, i.e., direct normalized generalized Foley-Sammon transform (DN-GFST) method in high dimensional spaces, is presented in this paper. It not only overcomes the limitation of traditional LDA that they overemphasize the larger distance between classes and cause large overlaps of neighboring classes, but also has the best separable ability in global sense. Lastly, our method is applied to facial image recognition, and the experimental results show that the performance of the present method is superior to those of the existing methods in terms of the classification error rate.