Bernadette Dorizzi, head of the project
Bio-IDentity federates the research teams from different Institut TELECOM's Telecommunication Schools, that are involved in technical and non technical aspects of Biometric Identity Verification. With a strong orientation on multimodality, this project aims at pursuing the research started on projects BIOMET I, II and BIOLAB, funded by Institut TELECOM, and ended in 2002 and 2003. Multidisciplinarity is the goal of Bio-IDentity, including researchers from Institut TELECOM in technical, sociological, juridical, deployment and market aspects.
In 2001-2002, the BIOMET research project (entitled "Multimodal Biometric Identity Verification"), funded by Institut TELECOM, aimed at using the complementarity of the different schools of Institut TELECOM involved in the area of identity verification, in order to build a securized system through the fusion of several modalities : signature, face, fingerprints, hand shape, and speech.
Indeed, several research teams were already strongly involved in Identity Verification or Pattern Recognition in different fields: Speech at ENST (Gérard Chollet) and EURECOM (Christian Wellekens), Handwriting at TELECOM SudParis (Bernadette Dorizzi), Face and Fingerprints at EURECOM (Jean-Luc Dugelay). Also, some teams worked on sensors design at TELECOM SudParis (infrared camera) and ENST (3D face acquisition system, by Charles Beumier and Francis Schmitt), and others on smart cards at ENST (Jean Leroux Les Jardins).
BIOMET's results were the following:
- operational verification systems for such modalities (on-line signature, speech, face, fingerprints, hand)
- a first multimodal database (5 modalities: face, speech, on-line handwritten signatures, fingerprints, hand shape) with time variability (92 individuals over 3 sessions) and novel sensors for the face modality
- validation of new approaches for multimodal fusion (signature and speech)
- several publications (IWFHR'2002, CIFED'2002, AVBPA'2003, WMUA'2003, RIVF'2003)
In 2003, the BIOLAB project started, a multidisciplinary project this time, motivated by the need of technological improvements on Biometrics on one hand and by a misknowledge of the different habit situations of the population on the other hand. The objective of this project was indeed the realization of an exploratory, multidisciplinary study allowing to consider some non technical aspects of Biometrics (juridical, sociological, market, deployment).
Bio-IDentity's Research Activities
Bio-IDentity at TELECOM SudParis
Technical Research Activities at TELECOM SudParis (Bernadette Dorizzi, Dijana Petrovska, Sonia Garcia-Salicetti, Mounim A. El Yacoubi, Anouar Mellakh, Walid Hizem, Nesma Houmani, Liliane Junie Nguepdjop, Viet Anh Phan)
On-line Signature Verification
Our team, strong of its experience in Handwriting Recognition (see Publications), works on Biometric Identity Verification by On-line Handwritten Signatures. We participated to the First International Signature Verification Competition, organized in Hong Kong in 2004 (SVC'2004). We have several publications in the field (see Publications). Bao Ly Van defended his PhD on December 2005 and a patent has been submitted
We work on Iris Identification in degraded situations with correlation and wavelet techniques. We are studying the HMM approach to iris verification. We are developping a common iris database with CASIA (Chinese Academy of Science Institute of Automation).
Face Recognition with TMSP's infrared camera prototype
We develop an original hardware solution for human face recognition. A differential CMOS
imaging system and a Synchronized flash camera have been developed to provide ambient light invariant images and facilitate segmentation of the face from the background. This invariance of face images demonstrated by our different prototype camera systems can result in a significant software/hardware simplification in biometric applications, especially on mobile platforms, in which the computation power and
memory capacity are both limited. In order to evaluate our prototypes we have built a face database with 4 different illumination conditions of 25 persons. These solutions give a significant improvement in performance using a simple correlation based algorithm associated with an adequate preprocessing. Finally, we have obtained promising results using fusion between different sensors.
2D face Verification
Principal Component Analysis has been used since 1990 [ in many recognition algorithms to get a face feature representation and to exploit the dimensionality reduction characteristic of the PCA.
The way to determine the optimal dimension of the reduced space, still not available. Another critical point when working with PCA is the influence of the training set, denoted here as PCA construction set.
We are working on the behaviour of the signal/residual information of the PCA eigenspectrum in order to determine an optimal threshold that could be used for the dimensionality reduction. We also point out the importance of having sets used for the PCA-eigenspace construction, by evaluating the baseline PCA face verification algorithm on the FRGCv2 database, using also images originally from BANCA database.
Multimodal Identity Verification
We work on multimodality from 3 perspectives :
Comparision between results obtained on virtual subjects multidatabases and real ones
Comparision between different rules for score fusion (SVM, normalized weighted summations)
Implementation of score fusion on a PDA with Three modalities (voice, face, signature)
3D Identity Verification
A PhD thesis in collaboration with Thales TRT aims at studiyng 3D and 2D features extraction.
Non Technical Research Activities at TELECOM SudParis
(Sylvie Craipeau, Claudine Guerrier, Gerard Dubey, Xavier Guchet)
Several social studies of biometric implementations are conducted in the framework of the french visa experimentation BIODEV and in the VINSI project (implementation of some biometric modalities on a mobile terminal).
The judicial work is performed through the Asphales project.