Fig 7 - uploaded by Jose Luis Alba-Castro
Content may be subject to copyright.
BIF with criterion A: effect of sweeping the number K of best selected features on system performance (TER(%) measures are reported) for the shape-driven, manual face-like mesh and rectangular grid methods in configurations I (left) and II (right)
Source publication
A comparative evaluation of two problems addressed in local Gabor feature-based face recognition is presented: localisation of points for feature extraction, and fusion of Gabor-based local similarity measures. For the former problem, three different point configurations are compared: a face-like mesh, a (rigid) rectangular grid and a shape-driven...
Contexts in source publication
Context 1
... BIF (Fig. 7): In configuration I, the best results are obtained with a number of features between 20 and 50. We should pay special attention to K ¼ 25 for the face-like mesh, since it achieves very competitive authentication rates (in fact, there is a range of values for K around 25 for which the TER is below 3%). In configuration II, and ...
Context 2
... performance of SFFS with criterion B is plotted in Fig. 10 for the shape-driven, face-like and rectangular meshes. By comparing these performances with those of Fig. 7, we realise that SFFS is outperformed by BIF. One would expect better performance from a method such as SFFS that actually takes interaction between features into account. Moreover, SFFS has empirically shown better per- formance than BIF in [29]. So, how do we explain the obtained results? To answer this question, we should remember ...
Context 3
... obtained using the global SFFS (Fig. 11, continue line and dots), is clearly worse than that of the SVM or MLP-AB. † Finally, we want to remark that client-specific BIF applied to the face-like mesh obtains comparable (and even better) results than more complex techniques, such as MLP-AB or SVM. Indeed, using K ¼ 25, the obtained TER measures (Fig. 7) are 2.43% (the lowest error rate among all tested techniques in configuration I) and 3% for configuration II. As discussed previously, BIF compensates its simplicity with the suitability to select client-specific locations therefore achieving low error rates. Analogously, AFS (Fig. 6) obtains competitive results specially for small ...
Similar publications
The benefits of applying automated fault detection and diagnosis (AFDD) to chillers include less expensive repairs, timely maintenance, and shorter downtimes. This study employs feature selection (FS) techniques, such as mutual-information-based filter and genetic-algorithm-based wrapper, to help search for the important sensors in data driven chil...
In this article, a variable selection procedure, called surrogate selection, is proposed which can be applied when a support vector machine or kernel Fisher discriminant analysis is used in a binary classification problem. Surrogate selection applies the lasso after substituting the kernel discriminant scores for the binary group labels, as well as...
Automatic Speaker Verification (ASV) systems are vulnerable to spoofing attacks. Most existing spoofing detection systems rely on two main points; the feature extraction and the classification methodology. In this paper, we propose a new strategy to recognize the veritable discourse from the spoofed one. The thought depends on the investigation of...
Citations
... Once demonstrated that GGs perform better in this scenario, we took advantage of the undelying statistics to compress data using coefficient quantization by means of Lloyd-Max algorithm. At this point, and in order to further reduce the template size, we decided to apply feature selection by means of the Best Individual Feature (BIF) algorithm [11,12,13]. This way, the template is compressed because of the lower number of features that are kept and, at the same time, system performance is drastically increased. ...
... The benefits of such methodology are not limited to reducing storage but also increasing system performance. One technique that has demonstrated good performance despite its simplicity is the Best Individual Feature (BIF) selection approach [11,12,13]. In [13], different tools for Gabor jet similarity fusion were evaluated, concludingTable 1. Face Verification on the XM2VTS database. ...
... One technique that has demonstrated good performance despite its simplicity is the Best Individual Feature (BIF) selection approach [11,12,13]. In [13], different tools for Gabor jet similarity fusion were evaluated, concludingTable 1. Face Verification on the XM2VTS database. False Acceptance Rate (FAR), False Rejection Rate (FRR) and Total Error Rate (TER) over the test set using both raw and compressed data and the whole set of 130 jets. ...
Gabor filters have demonstrated their effectiveness in auto- matic face recognition. However, one drawback of Gabor-based face rep- resentations is the huge amount of data that must be stored. One way to reduce space is to quantize Gabor coefficients using an accurate statisti- cal model which should reflect the behavior of the data. Statistical im- age analysis has revealed one interesting property: the non-Gaussianity of marginal statistics when observed in a transformed domain (like Discrete Cosine Transform, wavelet decomposition, etc.). Two models that have been used to characterize this non-normal behavior are the Generalized Gaussian (GG) and the Bessel K Form densities. This paper provides an empirical comparison of both statistical models in the specific scenario of modeling Gabor coefficients extracted from face images. Moreover, an application for biometric template reduction is presented: based on the underlying statistics, compression is first achieved via Lloyd-Max algo- rithm. Afterwards, only the best nodes of the grid are preserved using a simple feature selection strategy. Templates are reduced to less than 2 Kbytes with drastical improvements in performance, as demonstrated on the XM2VTS database.
We have developed a near-real-time computer system that can locate and track a subject's head, and then recognize the person by comparing characteristics of the face to those of known individuals. The computational approach taken in this system is motivated by both physiology and information theory, as well as by the practical requirements of near-real-time performance and accuracy. Our approach treats the face recognition problem as an intrinsically two-dimensional (2-D) recognition problem rather than requiring recovery of three-dimensional geometry, taking advantage of the fact that faces are normally upright and thus may be described by a small set of 2-D characteristic views. The system functions by projecting face images onto a feature space that spans the significant variations among known face images. The significant features are known as "eigenfaces," because they are the eigenvectors (principal components) of the set of faces; they do not necessarily correspond to features such as eyes, ears, and noses. The projection operation characterizes an individual face by a weighted sum of the eigenface features, and so to recognize a particular face it is necessary only to compare these weights to those of known individuals. Some particular advantages of our approach are that it provides for the ability to learn and later recognize new faces in an unsupervised manner, and that it is easy to implement using a neural network architecture.
This paper describes a system for multiple-object recognition and segmentation that (1) correctly identifies objects in a natural scene and provides a boundary for each object, (2) can identify multiple occurrences of the same object (e.g., two identical objects, side-by-side) in the scene from different training views. The algorithm is novel in that it employs statistical modeling to efficiently prune features from an identified object from the scene without disturbing similar features elsewhere in the scene. The originality of the approach allows one to analyze complex scenes that occur in nature contain multiple instances of the same object.
For cancers to develop, sustain and spread, the appropriation of key homeostatic physiological systems that influence cell growth, migration and death, as well as inflammation and the expansion of vascular networks are required. There is accumulating molecular and in vivo evidence to indicate that the expression and actions of the renin-angiotensin system (RAS) influence malignancy and also predict that RAS inhibitors, which are currently used to treat hypertension and cardiovascular disease, might augment cancer therapies. To appreciate this potential hegemony of the RAS in cancer, an expanded comprehension of the cellular actions of this system is needed, as well as a greater focus on translational and in vivo research.
Breast cancer patients have benefited from the use of targeted therapies directed at specific molecular alterations. To identify additional opportunities for targeted therapy, we searched for genes with marked overexpression in subsets of tumors across a panel of breast cancer profiling studies comprising 3,200 microarray experiments. In addition to prioritizing ERBB2, we found AGTR1, the angiotensin II receptor type I, to be markedly overexpressed in 10-20% of breast cancer cases across multiple independent patient cohorts. Validation experiments confirmed that AGTR1 is highly overexpressed, in several cases more than 100-fold. AGTR1 overexpression was restricted to estrogen receptor-positive tumors and was mutually exclusive with ERBB2 overexpression across all samples. Ectopic overexpression of AGTR1 in primary mammary epithelial cells, combined with angiotensin II stimulation, led to a highly invasive phenotype that was attenuated by the AGTR1 antagonist losartan. Similarly, losartan reduced tumor growth by 30% in AGTR1-positive breast cancer xenografts. Taken together, these observations indicate that marked AGTR1 overexpression defines a subpopulation of ER-positive, ERBB2-negative breast cancer that may benefit from targeted therapy with AGTR1 antagonists, such as losartan.
