Search

Automatic sign language recognition (ASLR) is quite a complex task, not only for the difficulty of dealing with very dynamic video information, but also because almost every sign language (SL) can be considered as an under-resourced language when it comes to language technology. Spanish sign language (LSE) is one of those under-resourced languages. Developing technology for SSL implies a number of technical challenges that must be tackled down in a structured and sequential manner. In this paper, some problems of machine-learning- based ASLR are addressed. A review of publicly available datasets is given and a new one is presented. It is also discussed the current annotations methods and annotation programs. In our review of existing datasets, our main conclusion is that there is a need for more with high-quality data and annotations.

Over the past few years, Presentation Attack Detection (PAD) has become a fundamental part of facial recognition systems. Although much effort has been devoted to anti-spoofing research, generalization in real scenarios remains a challenge. In this paper we present a new open-source evaluation framework to study the generalization capacity of face PAD methods, coined here as face-GPAD. This framework facilitates the creation of new protocols focused on the generalization problem establishing fair procedures of evaluation and comparison between PAD solutions. We also introduce a large aggregated and categorized dataset to address the problem of incompatibility between publicly available datasets. Finally, we propose a benchmark adding two novel evaluation protocols: one for measuring the effect introduced by the variations in face resolution, and the second for evaluating the influence of adversarial operating conditions., 8 pages, to appear at International Conference on Biometrics (ICB19)

In the current context of digital transformation, the increasing trend in the use of personal devices for accessing online services has fostered the necessity of secure cyberphysical solutions. Biometric technologies for mobile devices, and face recognition specifically, have emerged as a secure and convenient approach. However, such a mobile scenario also brings some specific threats, and spoofing attack detection is, without any doubt, one of the most challenging. Although much effort has been devoted in anti-spoofing techniques over the past few years, there are still many challenges to be solved when implementing these systems in real use cases. This chapter analyses some of the gaps between research and real scenario deployments, including generalisation, usability, and performance. More specifically, we will focus on how to select and configure an algorithm for real scenario deployments, paying special attention to use cases involving limited processing capacity devices (e.g., mobile devices), and we will present a publicly available evaluation framework for this purpose.

In this work we thoroughly study the well-known face verification Resnet model in dlib’s library to uncover inner features related to soft biometrics attributes like gender, race and age. The study makes use of the t-SNE technique to understand the evolution of clustering through the pretrained network layers and reveals an interesting property of t-SNE to spot separability of clusters in the original space. The performance of simple classifiers for the secondary soft-biometrics tasks through the network reinforce the findings about t-SNE. This study is extensible to any model that maps the input classes into an embedded low-dimensional space that learned to cluster them in task-meaningful sets. We conclude that a state of the art face verification model can be easily leveraged to state of the art soft biometrics model without resorting to fine-tuning convolutional weights, which also allows reducing the model size and inference time.

Developing robust and universal methods for unsupervised segmentation of moving objects in video sequences has proved to be a hard and challenging task that has attracted the attention of many researchers over the last decades. State-of-the-art methods are, in general, computationally heavy preventing their use in real-time applications. This research addresses this problem by proposing a robust and computationally efficient method, coined BMOG, that significantly boosts the performance of a widely used method based on a Mixture of Gaussians. The proposed solution explores a novel classification mechanism that combines color space discrimination capabilities with hysteresis and a dynamic learning rate for background model update. The complexity of BMOG is kept low, proving its suitability for real-time applications. BMOG was objectively evaluated using the ChangeDetection.net 2014 benchmark. An exhaustive set of experiments was conducted, and a detailed analysis of the results, using two complementary types of metrics, revealed that BMOG achieves an excellent compromise in performance versus complexity.

Developing robust and universal methods for unsupervised segmentation of moving objects in video sequences has proved to be a hard and challenging task. State-of-the-art methods show good performance in a wide range of situations, but systematically fail when facing more challenging scenarios. Lately, a number of image processing modules inspired in biological models of the human visual system have been explored in different areas of application. This paper proposes a bio-inspired boosting method to address the problem of unsupervised segmentation of moving objects in video that shows the ability to overcome some of the limitations of widely used state-of-the-art methods. An exhaustive set of experiments was conducted and a detailed analysis of the results, using different metrics, revealed that this boosting is more significant when challenging scenarios are faced and state-of-the-art methods tend to fail.

In this paper, we propose a different insight to analyze AdaBoost. This analysis reveals that, beyond some preconceptions, AdaBoost can be directly used as an asymmetric learning algorithm, preserving all its theoretical properties. A novel class-conditional description of AdaBoost, which models the actual asymmetric behavior of the algorithm, is presented.

Data security and privacy are crucial issues to be addressed for assuring a successful deployment of biometrics-based recognition systems in real life applications. In this paper, a template protection scheme exploiting the properties of universal background models, eigen-user spaces, and the fuzzy commitment cryptographic protocol is presented. A detailed discussion on the security and information leakage of the proposed template protection system is given. The effectiveness of the proposed approach is investigated with application to online signature recognition. The given experimental results, evaluated on the public MCYT signature database, show that the proposed system can guarantee competitive recognition accuracy while providing protection to the employed biometric data.

We present a solution for the development of multiplatform and web-oriented Java applications for biometric authentication based on the BioAPI framework. Our proposal is a single Java Native Interface wrapper that is compatible with the BioAPI instantiations of Windows and Linux/Unix operating systems. Following a study of existing wrappers, we extended the open-source Linux/Unix wrapper to include Windows support. In order to show the utility of this multiplatform Java wrapper, we discuss some of its possible uses and present a real-life application.

Large-scale deployment of biometric systems for web-based services has to tackle technological issues related to security, interoperability and accuracy, and social issues related to privacy protection and biometric acquisition process acceptance. The variety of biometric traits, capturing devices, targeted populations and working scenarios makes the development of a universal solution for all-purpose deployment quite a difficult task. This paper describes the design, implementation and applicability of an open framework for distributed biometric authentication oriented to access-control in web environment. The open principle makes this framework a novel and practical development tool for testing and integrating biometric algorithms and devices from third parties. Special attention has been paid to security and interoperability standards to ease concurrent integration and testing of biometric trait matchers developed by different laboratories or companies. Finally, in order to demonstrate the versatility and usability of the framework, we describe the construction process for a distributed multibiometric database acquisition tool based on this framework.

Predicting human emotions is catching the attention of many research areas, which demand accurate predictions in uncontrolled scenarios. Despite this attractiveness, designed systems for emotion detection are far off being as accurate as desired. Two of the typical measurements in human emotions are described in terms of the dimensions valence and arousal, which shape the Russell's circumplex where complex emotions lie. Thus, the Affect Recognition Sub-Challenge (ASC) of the third AudioVisual Emotion and Depression Challenge, AVEC'13, is focused on estimating these two dimensions. This paper presents a three-level fusion system combining single regression results from audio and visual features, in order to maximize the mean average correlation on both dimensions. Five sets of features are extracted (three for audio and two for video), and they are merged following an iterative process. Results show how this fusion outperforms the baseline method for the challenge database.

User's privacy protection is a very important issue which can seriously affect the usability of a biometric system, and prevent its successful establishment. In order to perform people verification while providing security and privacy to the employed characteristics, several biometric template protection schemes have been recently proposed. Unfortunately, when deploying a biometric cryptosystem, a significant intra-class variability, combined with the lack of enrollment data for estimating the templates' statistics, may prevent the users' verification due to the limits on error correction capability of the employed codes. In this paper we propose a feature-level fusion technique which can significantly reduce the intra-class variability of feature-based templates, thus allowing to reach low false recognition rates in protected systems. The benefits of the proposed approach are evaluated over a novel video-based face verification system relying on Universal Background Models and adapted user model eigen-projections. Tests over the BANCA database show good performance of the proposed features for face verification, and improved template protection when the proposed feature fusion approach is applied.

Gender classification is one of the most important tasks in automated face analysis, and has attracted the interest of researchers for years. Up to now, most gender classification approaches have been tested using single-database experiments, and on quite controlled datasets such as the FERET database, which are not representative of real world settings. However, a recent trend towards more realistic benchmarks has emerged within the face analysis community, leading to the appearance of databases and protocols such as the Labeled Faces in the Wild (LFW) database, and the so-called Gallagher's database, which comprises images collected from Flickr.

International audience; In this paper, we present the main results of the BioSecure Signature Evaluation Campaign (ESRA'2011). The objective of ESRA'2011 is to evaluate through two different tasks the resistance of different online signature systems to skilled forgeries categorized automatically according to their quality. Task 1 aims at studying with only coordinate time functions the influence of acquisition conditions (digitizing tablet vs. PDA) on systems' performance. The two BioSecure Data Sets DS2 and DS3 make this possible, since they contain data from the same 382 people, acquired respectively on a digitizer and on a PDA. Task 2 then aims at assessing the contribution of the five time functions available on a digitizer (coordinates, pressure, pen inclination) on systems' resistance to different qualities of skilled forgeries. Results of the 13 systems involved in this competition are reported and analyzed for both tasks in this paper. We observe that the best system in terms of performance on forgeries of "bad" quality is not necessarily the most resistant to an increased quality of skilled forgeries. Also, we note that mobile conditions are still threatening independently of the quality of forgeries. Finally, when adding pen inclination time functions to pressure and coordinates, we find that the gap between systems in terms of performance is wider than when only pen coordinates and pressure are considered

Cloud computing has achieved maturity, and there is a heterogeneous group of providers and cloud-based services. However, significant attention remains focused on security concerns. In many cases, security and privacy issues are a significant barrier to user acceptance of cloud computing systems and the advantages these offer with respect to previous systems. Biometric technologies are becoming the key aspect of a wide range of secure identification and personal verification solutions, but in a cloud computing environment they present some problems related to the management of biometric data, due to privacy regulations and the need to trust cloud providers. To overcome those problems in this paper, we propose a cryptobiometric system applied to cloud computing in which no private biometric data are exposed.

The protection of the templates stored in a biometric recognition system represents an issue of paramount importance for the security and privacy of the enrolled users, and directly affects the successful deployment of the system itself. In this paper we propose a protected on-line signature recognition system where the properties of Universal Background Models are exploited to provide a small dimensionality and a limited intra-class variability signature representation. The reported experimental results show that the employed signature representation and protection scheme allow to reach high recognition accuracy while providing protection to the considered biometric data.

In this paper we present a solution to obtain useful and reliable user logs in a Learning Management System (LMS).Current LMS logs are combined with biometric-based logs that show the student behaviour. Our system models the student behaviour, allowing to know exactly how much time the student spends in front of the computer examining the con-tents of the LMS. Besides, user verification and face tracking are also integrated, what guarantees that the student is the person actually interacting with the system. The presented multimodal solution for user tracking and user ver-ification combines face tracking, face verification, speakerverification and fingerprint verification. Face tracking and face verification are performed in a non-collaborative fashion. Fingerprint or speaker verification is performed on demand, with the aim of avoiding a negative in?uence of adverse environmental or behavioural human factors in the reliability of the user logs generated by the system. These circumstances can thwart the non collaborative face veri-fication performance involved in the tracking process. The presented solution solves the problem of user tracking and authentication even in adverse environments for face verifi-cation.

This paper addresses the problem of biometric audiovisual person authentication in realistic acquisition conditions. Differences in environmental factors or acquisition devices between enrollment and test conditions modify the verification scores distribution and degrade verification performance if not taken into account. A theoretical framework that incorporates quality measures to biometric authentication is introduced. As a result, the necessary and sufficient condition that a quality measure must hold to enable improved verification performance is given. Two quality-based score normalization techniques are derived that successfully incorporate quality factors in the verification decision process. Experiments on the multimodal BANCA database for video-based face verification and speaker verification show a statistically significant verification performance improvement when using the proposed quality-based score normalization techniques.

Beside the optimization of biometric error rates the overall security system performance in respect to intentional security attacks plays an important role for biometric enabled authentication schemes. As traditionally most user authentication schemes are knowledge and/or possession based, firstly in this paper we present a methodology for a security analysis of Internet-based biometric authentication systems by enhancing known methodologies such as the CERT attack-taxonomy with a more detailed view on the OSI-Model. Secondly as proof of concept, the guidelines extracted from this methodology are strictly applied to an open source Internet-based biometric authentication system (BioWebAuth). As case studies, two exemplary attacks, based on the found security leaks, are investigated and the attack performance is presented to show that during the biometric authentication schemes beside biometric error performance tuning also security issues need to be addressed. Finally, some design recommendations are given in order to ensure a minimum security level.

This paper addresses the incorporation of quality measures to video-based person authentication. A theoretical framework to incorporate quality measures in biometric authentication is exposed. Two different quality-based score normalization techniques are derived from this theoretical framework. Furthermore, a quality-based frame selection technique and a new face image quality measure are also presented. The ability of this quality measure and the proposed quality-based score normalization techniques and quality-based frame selection technique to improve verification performance is experimentally evaluated in a video-based face verification experiment on the BANCA Database.

Despite of the advances in the e-learning domain during the last decades, there is a lack of suitable mechanism to carry out assessment with the appropriate measures to avoid cheating. Current LMSs do not provide the needed features to check that the intended student is taking the online exam by himself, or even to know if he has spent the whole session time in front of the computer. This paper presents a web-based application that offers biometric authentication based on face-recognition. This application, which can be easily integrated in currently available LMSs, is able to use face-recognition during access control, tracking and assessment. Thus it is possible to improve security during critical phases in the learning process (e.g. assessment).

Preference learning has recently received a lot of attention within the machine learning field, concretely learning by pairwise comparisons is a well-established technique in this field. We focus on the problem of learning the overall preference weights of a set of alternatives from the (possibly conflicting) uncertain and imprecise information given by a group of experts into the form of interval pairwise comparison matrices. Because of the complexity of real world problems, incomplete information or knowledge and different patterns of the experts, interval data provide a flexible framework to account uncertainty and imprecision. In this context, we propose a two-stage method in a distance-based framework, where the impact of the data certainty degree is captured. First, it is obtained the group preference matrix that best reflects imprecise information given by the experts. Then, the crisp preference weights and the associated ranking of the alternatives are derived from the obtained group matrix. The proposed methodology is made operational by using an Interval Goal Programming formulation.

The pairwise comparison method is an interesting technique for assessing priority weights for a finite set of objects. In fact, some web search engines use this inference tool to quantify the importance of a set of web sites. In this paper we deal with the problem of incomplete paired comparisons. Specifically, we focus on the problem of retrieving preference information (as priority weights) from incomplete pairwise comparison matrices generated during a group decision-making process. The proposed methodology solves two problems simultaneously: the problem of deriving preference weights when not all data are available and the implicit consensus problem. We consider an approximation methodology within a flexible and general distance framework for this purpose.

In this paper we introduce an adaptive image segmentation neural network based on a Gaussian mixture classifier that is able to accommodate unlabeled data in the training process to improve generalization when labeled data is insufficient. The classifier is trained by maximizing the joint-likelihood of features and labels over all the data set (labeled and unlabeled). The classifier builds grey- level images with estimation of class-posteriors (as many images as classes) that feed the segmentation algorithm. The paper is focused on the adaptive classification part of the algorithm. The classification tests are performed over Landsat TM mini-scenes. We assess the efficiency of the adaptive classifier depending on the model complexity and the proportion of labeled/unlabeled data.