Hidden Conditional Random Fields for Face Recognition

This paper proposes a hidden conditional random field(HCRF) model for face recognition. Face images are separated as a series of block and 2D-DCT feature vectors is extracted in each block. Libsvm is used as a local discriminative model that outputs the association of the feature vectors with latent variables. HCRF is used to model the entire hidden state sequence. The method proposed in this paper achieves a higher recognition rate compared to the state-of-the-art in ORL database. The resusts indicate that integrating various dependencies between latent variables is useful for face recognition.