Face recognition based on subject dependent Hidden Markov Models

In this paper we present an automatic face recognition system based on incremental Singular Values Decomposition (SVD) and subject dependent Hidden Markov Models (HMM). For each subject, an individual HMM is trained with features, extracted from the orthogonal decomposition (SVD) of the subject's training images. The main advantage of the proposed SVD-HMM recognition system is the robustness against image dimensionality reduction. The system was tested on two benchmark face datasets - the Olivetti Research Laboratory (ORL) and the YALE database. The SVD-HMM was further compared with a standard SVD face recognition. SVD applied to the original (full size) images performs similarly to the SVD-HMM applied to the compressed (half of the original size) images. SVD degrades rapidly when the image is compressed.

I. Introduction

Face Recognition is an active research area with increasing number of applications such as in security, robotics, human-computer-interfaces, games, entertainment, [1]–[3]. Typical face recognition methods are based on PCA eigenfaces, Linear Discriminate Analysis (LDA), Singular Values Decomposition (SVD), [4]. The SVD advantage is that it can adjust the variations that are present in the local statistics of an image. In [5] a novel technique for image representation is proposed where the SVD is incorporated to a hierarchical decomposition named Inverse Difference Pyramid (IDP). This approach is developed by analogy with the hypothesis for the way humans do image recognition using consecutive approximations with increasing similarity, [6]. In [7], a face recognition system based on Hidden Markov Models (HMM) is proposed. HMM is a statistical model for modeling sequential or time series data that has been successfully applied in many problems such as speech and character recognition, robot control, information extraction from text data [8], [9]. The present paper is an extension of our previous results [10], where we introduced the SVD-HMM. Here we go further and assess the sensitivity of the recognition system with respect to variations in meta parameters and compare SVD-HMM with the SVD-based face detection. The paper is organized as follows. In section 2 we describe the HMM structure. The incremental SDV-HMM face recognition is presented in section 3. The datasets and image preprocessing are introduced in section 4. The implementation, experiments and results are discussed in section 5. Finally our conclusions are outlined in section 6.