Feature Extraction Method Based on the Generalised Fisher Discriminant Criterion and Facial Recognition

In this paper we view the optimal set of discriminant vectors as a global transform, and consider its separability from a global view point. Based on this idea, the concept of a generalised Fisher discriminant criterion and that of a generalised optimal set of discriminant vectors is introduced. After that, a new algorithm is given to calculate the generalised optimal set of discriminant vectors defined in this paper, which is particularly suited to the case of a small number of samples where the scatter matrix is singular. It is then applied to an experiment on human facial recognition, and the results show that the new algorithm is superior to existing methods in terms of correct classification rate.