A Novel Discriminant Non-Negative Matrix Factorization Algorithm With Applications to Facial Image Characterization Problems

Non-negative matrix factorization (NMF) is a powerful approach to single channel audio source separation. In a supervised setting, NMF is first applied to train the basis spectra of each sound source. At test time, NMF is applied to the spectrogram of a mixture signal using the pretrained spectra. The source signals can then be separated out using a Wiener filter. A typical way to train the basis spectra of each source is to minimize the objective function of NMF. However, the basis spectra obtained in this way do not ensure that the separated signal will be optimal at test time due to the inconsistency between the objective functions for training and separation (Wiener filtering). To address this, a framework called discriminative NMF (DNMF) has recently been proposed. In in this work a multiplicative update algorithm was proposed for the basis training, however one drawback is that the convergence is not guaranteed. To overcome this drawback, this paper proposes using a majorization-minimization principle to develop a convergence-guaranteed algorithm for DNMF. Experimental results showed that the proposed algorithm outperformed standard NMF and DNMF using a multiplicative update algorithm as regards both the signal-to-distortion and signal-to-interference ratios.

Discriminative nonnegative matrix factorization (DNMF) has been investigated as a promising basis-learning method for monaural source separation. To the best of our knowledge, however, no good and sound discussion has been made on quantitative definition of discriminability and it is difficult to evaluate how discriminative DNMF is actually. This paper introduces a quantitative measure to calculate how discriminative two NMF bases are. From the viewpoint of our measure, we compare three basis-learning methods of plain NMF, DNMF, and minimum-volume (min-vol) NMF. Experimental results of monaural speech separation reveal that min-vol NMF actually learns as discriminative bases as DNMF and achieves the best separation performance. This is probably because min-vol NMF can learn the most compact basis possible that can cover training data.

Recent year, the method based on face verification has been one of the most vibrant study aspect in face verification field. The problem of face verification can be defined: imputing the picture or video, identifying or validating by using face database. video-video face verification means inputting video setting, verification by using face video data。It has most useful information, including: a lots of pictures of one person, continuity of face in time and space in video, three dimension face model etc. 1.PCA face verification Recent year, the method based on face verification has been one of the most vibrant study aspect in face verification field. The problem of face verification can be defined: imputing the picture or video, identifying or validating by using face database. video-video face verification means inputing video setting, verification by using face video data。It has most useful information, including: a lots of pictures of one person, continuity of face in time and space in video, three dimension face model etc. The way that face information be described in video in present literature can be summarized as follows: vector、matrix probability、 dynamic model 、fashion. daopting modes of probability and fashion need mass swatch which reflect distribution of face, can accurately describe face distribution. Dynamic model can better use time and space information , but the method is compicated, calculation is great. The describe way of imputing vector has a big defect, the randomicity of choose swatch.the way of matrix is comparatively simple, can apply the informatin of discontinuous pictures in time,but how to depict relationship between matrix is a worthful contect. Face character based on PCA (Principal Component Analysis, PCA)is the classic arithmetic in face verification, it is a face face veification and describe technology educe from bases analysis. Face character is a simple 、fast、 useful、 arithmetic based on transform coefficient character. Face verification: compare the given face portrait with the picture stored in computer, determine the face is whether the appointed guy, it is a one to one matching process, usually one person stored more face portrait in different angle in computer. A face verification system based on video including face detect model face portrait pretreatment model face character distill model and face verification model. 1.1PCA face verification Face verification: compare the given face portrait with the picture stored in computer, determine the face is whether the appointed guy, it is a one to one matching process, usually one person stored more face portrait in different angle in computer. A face verification system (3) based on video including face detect model face portrait pretreatment model face character distill model and face verification model,as picture 1.

Non-negative Matrix Factorization (NMF) is among the most popular subspace methods widely used in a variety of pattern recognition applications. Recently, a discriminant NMF method that incorporates Linear Discriminant Analysis criteria and achieves an efficient decomposition of the provided data to its salient parts has been proposed. An extension of this work specialized for classification, optimized using projected gradients in order to ensure converge to a stationary limit point, resulted in a more efficient method of the latter approach. Assuming multimodality of the underlying data samples distribution and incorporating clustering discriminant inspired constraints into the NMF decomposition cost function, resulted in the Subclass Discriminant NMF algorithm which found to outperform both approaches under real life settings. In this work we review all these methods in the context of various pattern recognition problems using facial images.

Facial expression recognition is a very important research field to understand human emotions. Many facial expression recognition systems have been proposed in the literature over the years. Some of these methods use neural network approaches with deep architectures to address the problem. Although it seems that the facial expression recognition problem has been solved, there is a large difference between the results achieved using the same database to train and test the network and the cross-database protocol. In this paper, we extensively investigate the performance influence of fine-tuning with cross-database approach. In order to perform the study, the VGG-Face Deep Convolutional Network model (pre-trained for face recognition) was fine-tuned to recognize facial expressions considering different well-established databases in the literature: CK+, JAFFE, MMI, RaFD, KDEF, BU3DFE, and AR Face. The cross-database experiments were organized so that one of the databases was separated as test set and the others as training, and each experiment was ran multiple times to ensure the results. Our results show a significant improvement on the use of pre-trained models against randomly initialized Convolutional Neural Networks on the facial expression recognition problem, for example achieving 88.58%, 67.03%, 85.97%, and 72.55% average accuracy testing in the CK+, MMI, RaFD, and KDEF, respectively. Additionally, in absolute terms, the results show an improvement in the literature for cross-database facial expression recognition with the use of pre-trained models.

Non-negative subspace feature representation for few-shot learning in medical imaging

The objective of single-channel source separation is to accurately recover source signals from mixtures. Non-negative matrix factorization (NMF) is a popular approach for this task, yet previous NMF approaches have not optimized directly this objective, despite some efforts in this direction. Our paper introduces discriminative training of the NMF basis functions such that, given the coefficients obtained on a mixture, a desired source is optimally recovered. We approach this optimization by generalizing the model to have separate analysis and reconstruction basis functions. This generalization frees us to optimize reconstruction objectives that incorporate the filtering step and SNR performance criteria. A novel multiplicative update algorithm is presented for the optimization of the reconstruction basis functions according to the proposed discriminative objective functions. Results on the 2nd CHiME Speech Separation and Recognition Challenge task indicate significant gains in source-to-distortion ratio with respect to sparse NMF, exemplar-based NMF, as well as a previously proposed discriminative NMF criterion.

In this paper, a novel supervised feature extraction method is presented. The method employs discriminant analysis in the features derived by non-negative matrix factorization (NMF). In this way, a two phase discriminant feature extraction procedure is implemented, namely NMF plus linear discriminant analysis (LDA). The introduced method has been applied to the problem of frontal face verification using the well known XM2VTS database, where a better performance than NMF, eigenfaces and Fisherfaces has been achieved

In this paper, the Direction basis function networks(DBFNs) are proposed to improve the problem of facial expression recognition. Due to the diversity and nonlinear characteristic of facial expression itself, We proved a new linear feature extraction approach of facial expression which achieved a good directional feature. Using Cohn-Kanade facial expression database, our method provides a better performance for facial expression recognition.

Most of the facial expression recognition methods consider that both training and testing data are equally distributed. As facial image sequences may contain information for heterogeneous sources, facial data may be asymmetrically distributed between training and testing, as it may be difficult to maintain the same quality and quantity of information. In this work, we present a novel classification method based on the learning using privileged information (LUPI) paradigm to address the problem of facial expression recognition. We introduce a probabilistic classification approach based on conditional random fields (CRFs) to indirectly propagate knowledge from privileged to regular feature space. Each feature space owns specific parameter settings, which are combined together through a Gaussian prior, to train the proposed t-CRF+ model and allow the different tasks to share parameters and improve classification performance. The proposed method is validated on two challenging and publicly available benchmarks on facial expression recognition and improved the state-of-the-art methods in the LUPI framework.

In this paper a novel non-linear subspace method for face verification is proposed. The problem of face verification is considered as a two-class problem (genuine versus impostor class). The typical Fisher's linear discriminant analysis (FLDA) gives only one or two projections in a two-class problem. This is a very strict limitation to the search of discriminant dimensions. As for the FLDA for N class problems (N is greater than two) the transformation is not person specific. In order to remedy these limitations of FLDA, exploit the individuality of human faces and take into consideration the fact that the distribution of facial images, under different viewpoints, illumination variations and facial expression is highly complex and non-linear, novel kernel discriminant algorithms are proposed. The new methods are tested in the face verification problem using the XM2VTS database where it is verified that they outperform other commonly used kernel approaches.

We discuss about a design process of 'personalized' classifier with soft computing techniques. Based on a human's way of thinking, a construction methodology for personalized classifier is mentioned. Here, two fuzzy similarity measures and ensemble of classifiers are effectively used. A feature selection method for the fuzzy neural networks plays also a key role for construction of personalized classifier. As one of the possible applications, facial expression recognition problem is discussed. The numerical result shows that the proposed method can enhance the classification rate for 22 persons from 81.0% to 90.4%.

Even though facial expressions have universal meaning in communications, their appearances show a large amount of variation due to many factors, such as different image acquisition setups, different ages, genders, and cultural backgrounds etc. Collecting enough amounts of annotated samples for each target domain is impractical, this paper investigates the problem of facial expression recognition in the more challenging situation, where the training and testing samples are taken from different domains. To address this problem, after observing the fact of unsatisfactory performance of the Kernel Mean Matching (KMM) algorithm, we propose a supervised extension that matches the distributions in a class-to-class manner, called Supervised Kernel Mean Matching (SKMM). The new approach stands out by taking into consideration both matching the distributions and preserving the discriminative information between classes at the same time. The extensive experimental studies on four cross-dataset facial expression recognition tasks show promising improvements of the proposed method, in which a small number of labeled samples guide the matching process.

Emotions are dynamic biological states that are connected to all of the nerve systems. The problem of facial expression recognition has been thoroughly investigated, leading to the development of some robust and accurate face recognition algorithms. The effectiveness of three such algorithms (CNN, VGG16, and ResNet50) that have been widely studied and applied in the research community are investigated and compared in this paper. The aim is to use grayscale images to train these training models and compare their accuracy and data losses. The system will be able to detect the seven facial expressions Angry, Neutral, Contempt, Disgust, Fear, Happy, and Sad after training these models. To compare their precision, the same batch size and epoch were used. After reviewing all possible evaluations based on these output matrices, it is clear that all three networks produce reliable effect identification, with CNN being the most accurate.

In this paper, we propose to solve the problem of facial expression recognition in videos by implementing a two-stage procedure, in which, firstly, facial features are extracted from all frames using an EfficientNet-based model. The latter is pre-trained to identify facial attributes and further fine-tuned on an external dataset for the emotion classification task. Secondly, multiple statistical functions are calculated and used in the aggregation process to create a single video representation. Furthermore, we propose a new technique for sequence, frame-level attention models, and 1D convolutions by concatenating the output of a statistical function with the facial features. It was experimentally shown that the proposed approach leads to state-of-the-art results on the AFEW 8.0 dataset.