Exploiting discriminant information in nonnegative matrix factorization with application to frontal face verification

Abstract

In this paper, two supervised methods for enhancing the classification accuracy of the Nonnegative Matrix Factorization (NMF) algorithm are presented. The idea is to extend the NMF algorithm in order to extract features that enforce not only the spatial locality, but also the separability between classes in a discriminant manner. The first method employs discriminant analysis in the features derived from NMF. In this way, a two phase discriminant feature extraction procedure is implemented, namely NMF plus Linear Discriminant Analysis (LDA). The second method incorporates the discriminant constraints inside the NMF decomposition. Thus, a decomposition of a face to its discriminant parts is obtained and new update rules for both the weights and the basis images are derived. The introduced methods have been applied to the problem of frontal face verification using the well known XM2VTS database. Both methods greatly enhance the performance of NMF for frontal face verification. Index TermsSubspace techniques, non-negative matrix factorization, linear discriminant analysis, frontal face verification.

Co-regularized multiview nonnegative matrix factorization with correlation constraint for representation learning

“…In recent years, many NMF-related feature extraction algorithms have been proposed. For example, LNMF [29], GNMF [4], DNMF [53], RDPNA [52] and so on. These algorithms can generate superior clustering results, but only deal with single-view data.…”

Locality-preserving nonnegative matrix factorization for hyperspectral image classification

“…To improve the classification performance and to obtain a more intuitive basis, we further exploit LFDA within the LP-NMF subspace in order to incorporate the local discriminant constraints within the LPNMF decomposition; the resulting technique is called LPNMF-LFDA-GMM. A similar idea of employing LDA in an NMF subspace was studied for facerecognition tasks in [10].…”

Enhancement of multimodal biometric segregation using unconstrained cohort normalisation

“…The data sets considered for the face and voice modalities in this investigation are extracted from the XM2VTS and TIMIT databases, respectively [17,19]. Using these biometric data sets, a total of 235 chimerical identities are The results for the verification experiments in this part of the study are presented as equal error rates (EERs) in Table 1.…”

“…The data sets considered for the face and voice modalities in this case are extracted from the XM2VTS (clean images) [17] and from the 1-speaker detection task of the NIST Speaker Recognition Evaluation 2003 (degraded speech) databases, respectively [12]. Using these data sets, again a total of 235 chimerical identities are formed.…”