Project

Back to overview

Embedding Local Quality Measures in Multimodal Biometric Recognition Systems

English title Embedding Local Quality Measures in Multimodal Biometric Recognition Systems
Applicant Drygajlo Andrzej
Number 146826
Funding scheme Project funding (Div. I-III)
Research institution Laboratoire de traitement des signaux 5 EPFL - STI - IEL - LTS5
Institution of higher education EPF Lausanne - EPFL
Main discipline Information Technology
Start/End 01.04.2013 - 31.03.2015
Approved amount 113'306.00
Show all

All Disciplines (2)

Discipline
Information Technology
Other disciplines of Engineering Sciences

Keywords (6)

quality; fingerprints; recognition; multimodality; iris; biometrics

Lay Summary (French)

Lead
Avec une augmentation de la fraude à l'identité, il y a un besoin croissant d'identifier les êtres humains automatiquement à la fois localement et à distance. L'acquisition de données biométriques d'une qualité suffisante et leur utilisation pour la prise de décision fiable est d'une importance critique pour les systèmes automatiques de reconnaissance de personne. Vérification d'identité de personne (authentification biométrique) est également une technologie multimodale à part entière.
Lay summary

En combinant plusieurs modalités, les systèmes biométriques multimodaux tiennent la promesse d’une authentification de personne flexible et robuste, en évitant l’exclusion ou la discrimination de personne. L’objectif de ce projet est d’élargir l’état de l’art des méthodes biométriques multimodales actuel en intégrant des mesures de qualité locale. Aborder la question de la qualité des données locales est vital pour les systèmes d’authentification biométrique multimodale avec descripteurs à caractère local modernes qui exploitent les données biométriques, habituellement localement affectées par les conditions extérieures et le comportement de l’utilisateur. Dans le présent projet, nous allons nous pencher sur le problème de la façon d’intégrer les mesures de qualité locale à la reconnaissance biométrique en utilisant les descripteurs à caractère local, afin d’obtenir des meilleures taux de reconnaissance, en particulier pour les empreintes digitales, l’iris et le visage. Nous allons également enquêter sur des nouvelles approches pour combiner les mesures de qualité locale et globale dans les systèmes biométriques multimodaux.

Direct link to Lay Summary Last update: 04.04.2013

Responsible applicant and co-applicants

Employees

Publications

Publication
How synthetic fingerprints can improve pre-selection of MCC pairs using local quality measures
Izadi Hamed, Drygajlo Andrzej (2015), How synthetic fingerprints can improve pre-selection of MCC pairs using local quality measures, in Biometrics and Forensics (IWBF), 2015 International Workshop on, Gjovik.
Palm vein recognition with local texture patterns
Mirmohamadsadeghi Leila, Drygajlo Andrzej (2014), Palm vein recognition with local texture patterns, in IET Biometrics, 3(4), 198-206.
Estimation of cylinder quality measures from quality maps for Minutia-Cylinder Code based latent fingerprint matching
Izadi Hamed, Drygajlo Andrzej (2013), Estimation of cylinder quality measures from quality maps for Minutia-Cylinder Code based latent fingerprint matching, in Proceedings of Biometric Technologies in Forensic Science, Nijmegen.
Quality and Reliability in Multimodal and Multi-classifier Biometric Person Recognition
Drygajlo Andrzej (2013), Quality and Reliability in Multimodal and Multi-classifier Biometric Person Recognition, in Popescu-Belis Andrei, Bourlard Hervé (ed.), EPFL Press, Lausanne, 189-204.
Speaker verification in score-ageing-quality classification space
Kelly Finnian, Drygajlo Andrzej, Harte Naomi (2013), Speaker verification in score-ageing-quality classification space, in Computer Speech and Language, 27(5), 1068-1084.
Discarding low quality Minutia Cylinder-Code pairs for improved fingerprint comparison
Izadi Hamed, Drygajlo Andrzej, Discarding low quality Minutia Cylinder-Code pairs for improved fingerprint comparison, in 2015 International Conference of the Biometrics Special Interest Group (BIOSIG), Darmstadt.

Collaboration

Group / person Country
Types of collaboration
COST IC1106 Belgium (Europe)
- in-depth/constructive exchanges on approaches, methods or results
- Research Infrastructure
- Exchange of personnel

Scientific events

Active participation

Title Type of contribution Title of article or contribution Date Place Persons involved
3rd International Workshop on Biometrics and Forensics Talk given at a conference How synthetic fingerprints can improve pre-selection of MCC pairs using local quality measures 03.03.2015 Gjøvik University College (GUC), Gjøvik, Norway Drygajlo Andrzej; Izadi Hamed Mohammad;
Biometric Technologies in Forensic Science Talk given at a conference Estimation of cylinder quality measures from quality maps for Minutia-Cylinder Code based latent fingerprint matching 14.10.2013 Nijmegen, Netherlands Izadi Hamed Mohammad;
14th Annual Conference of the
International Speech Communication Association (Interspeech 2013) Talk given at a conference Forensic Automatic Speaker Recognition: Theory, Implementation and Practice 25.08.2013 Lyon, France Drygajlo Andrzej;


Knowledge transfer events

Active participation

Title Type of contribution Date Place Persons involved


Associated projects

Number Title Start Funding scheme
118049 ABID2: Applying Biometrics to Identity Documents 01.10.2007 Project funding (Div. I-III)
127321 Embedding Quality and Reliability in Multimodal, Multiclassifier Biometric Recognition Systems 01.11.2010 Project funding (Div. I-III)

Abstract

The objective of this project is to extend the existing state-of-the-art uni- and multi-modal biometric identity verification and identification methods to incorporate local quality measures. Addressing the issue of data quality is vital for automated multimodal biometric authentication systems that operate on biometric data, usually affected by external conditions and user behavior. Acquiring biometric data of sufficient quality and suitability and using it for reliable decision-making is of critical importance for automatic person recognition systems.With an increase in identity fraud and the emphasis on security, there is a growing and urgent need to identify humans both locally and remotely on an automatic routine basis. The travel identity documents (e-passports) deployed on very large scale and the increasing use of information sensitive applications, such as e-government, e-commerce, e-banking, and e-health, has triggered a real need for reliable and widely acceptable automated control mechanisms for checking the identity of an individual. Biometrics technologies, using one or more of a person’s distinct behavioral or physiological characteristics, appears as a viable alternative to more traditional approaches such as identity badges, passwords and PIN codes.Biometric person identity verification (biometric authentication) is a multimodal technology in its own right, with many potential applications. Every biometric modality has some limitations, e.g., a biometric system using single modality may not be able to acquire meaningful biometric data from a subset of users, for example visually handicapped or disabled people. One possible solution to these problems is the endemic use of multiple biometrics. Multimodal biometric systems hold the promise of flexible and robust person authentication avoiding person exclusion or discrimination. The demand for multimodal biometric systems is increasing dramatically, due to security pressures and the need of successful deployment of such systems worldwide. However, our recent studies in international frameworks (e.g. Biosecure Network of Excellence, COST 2101 "Biometrics for Identity Documents and Smart Cards" and COST 1106 "Integrating Biometrics and Forensics for the Digital Age") demonstrate how performance of biometric systems is heavily affected by the quality of biometric samples.Like any other pattern recognition system, biometric recognition systems use either straightforward global features, which normally describe the whole biometric sample as a feature vector, or local features, which represent local parts (or points) within the sample, being more reliable regarding local variations. In modern local feature extraction techniques, some local feature descriptors are built to encode the relationship between the initially extracted local features within a local area in terms of some measures invariant to global transformations such as rotation, translation and scaling. Local feature descriptors are becoming increasingly popular in biometrics as a means for fast local comparisons.The state-of-the-art automatic biometric recognition systems are still prone to produce errors. Therefore, one of the most important goals in biometrics research should be reducing such errors. Errors can however come from different sources such as noise and artifacts in biometric samples, non-uniqueness of taken samples, sensing conditions, sensor characteristics, not well chosen feature sets, and unreliable classifiers. These factors cause degradation in biometric sample quality and this would ultimately matter in the ability of the biometric systems for correct recognition. Therefore, the quality of samples can be taken into account in order to improve the recognition rates. That is why many works has been done over the past decade to introduce some quantitative measures of quality for biometric samples. Here, by quality measure, we basically mean the degree of being free from corrupting degradations. Local quality measures represent the quality of local parts within the biometric sample versus the global quality measure which is a measure for quality of the whole sample.Incorporating local quality measures in biometric recognition systems based on local feature descriptors is a challenging research problem, which has not yet been studied widely. In general, in the present project we will address the problem of how to embed local quality measures in matching using modern local descriptors, in order to obtain better recognition rates, particularly for fingerprint, iris and face modalities. Then, we will investigate the novel approaches for multimodal biometric fusion integrating both local and global quality measures, taking into account the state-of-the-art models already available for incorporating global quality measures. Within the two year period of the project the scientific and technological objectives are addressed along the following innovative axes of research and development:•New local quality measures for fingerprint, iris and face modalities and novel frameworks for embedding them in biometric recognition, especially in presence of modern local feature descriptors.•New, universal data-driven models and processing techniques for multimodal robust biometric recognition systems integrating both local and global quality measures.The present proposal results from the main research area of the EPFL-LIDIAP Speech Processing and Biometrics Group (GTPB) and is a logical continuation of research of the ongoing SNSF project “Embedding Quality and Reliability in Multimodal, Multiclassifier Biometric Recognition Systems”, foreseen in the proposal of that project. The present project also will continue fundamental research activities started within the European Science Foundation Action COST 2101 “Biometrics for Identity Documents and Smart Cards” chaired by Dr. Drygajlo, and is in accordance with new European Science Foundation Action COST IC1106 “Integrating Biometrics and Forensics for the Digital Age”, as well as the development activities of two Swiss companies (Covadis and IrisGuard in Geneva), the group participates in.
-