Generalized discriminant analysis model and its extension for facial expression recognition

Over the past few decades, numerous linear discriminant analysis based extensions are proposed for dimensionality reduction. However, most of them are developed intuitively according to specific motivations by employing various techniques. Therefore, it will be instructive to provide a unified discriminant analysis model for exploring the commonalities and differences. In this paper, we propose a generalized discriminant analysis model (GDA model) by comprehensively considering the key components in designing a discriminator in which various LDA-based methods can be unified into this framework. Thus we can get better understanding of the inherent relationship among these algorithms by interpreting them from a unified perspective.