### 1A User-specific and Selective Multimodal Biometric Fusion Strategy by Ranking Subjects

"... The recognition performance of a biometric system varies significantly from one enrolled user to another. As a result, there is a need to tailor the system to each user. This study investigates a relatively new fusion strategy that is both user-specific and selective. By user-specific, we understand ..."

Abstract
- Add to MetaCart

(Show Context)
The recognition performance of a biometric system varies significantly from one enrolled user to another. As a result, there is a need to tailor the system to each user. This study investigates a relatively new fusion strategy that is both user-specific and selective. By user-specific, we understand that each user in a biometric system has a different set of fusion parameters that have been tuned specifically to a given enrolled user. By selective, we mean that only a subset of modalities may be chosen for fusion. The rationale for this is that if one biometric modality is sufficiently good to recognize a user, fusion by multimodal biometrics would not be necessary, in principle. The technical challenge here is that there exists no criterion In practice, however, We advance the state of the art in user-specific and selective fusion in the following ways: (1) provide thorough analyses of (a) the effect of pre-processing the biometric output (prior to applying a user-specific score normalization procedure) in order to improve its central tendency and (b) the generalisation ability of user-specific parameters; (2) propose a criterion to rank the users based solely on a training score dataset in such a way that the obtained rank order will maximally correlate with the rank order that is obtained if it were to be computed on the test set; and, (3) experimentally demonstrate the performance gain of a user-specific and-selective fusion strategy across fusion data sets and different “pruning rate ” that controls the percentage of subjects for whom fusion is not required. 15 sets of multimodal fusion experiments carried out

### An Investigation of F-ratio Client-Dependent Normalisation on Biometric Authentication Tasks

, 2004

"... submitted for publication ..."

(Show Context)
### Normalisation on Biometric Authentication Tasks

, 2004

"... submitted for publication ..."

(Show Context)
### How Do Correlation and Variance of Base-Experts Affect Fusion in Biometric Authentication Tasks?

, 2004

"... Abstract. Combining multiple information sources such as subbands, streams (with different features) and multi modal data has shown to be a very promising trend, both in experiments and to some extend in real-life biometric authentication applications. Despite considerable efforts in fusions, there ..."

Abstract
- Add to MetaCart

(Show Context)
Abstract. Combining multiple information sources such as subbands, streams (with different features) and multi modal data has shown to be a very promising trend, both in experiments and to some extend in real-life biometric authentication applications. Despite considerable efforts in fusions, there is a lack of understanding on the roles and effects of correlation and variance (of both the client and impostor scores of base-classifiers/experts). Often, scores are assumed to be independent. In this paper, we explicitly consider this factor using a theoretical model, called Variance Reduction-Equal Error Rate (VR-EER) analysis. Assuming that client and impostor scores are approximately Gaussian distributed, we showed that Equal Error Rate (EER) can be modeled as a function of F-ratio, which itself is a function of 1) correlation, 2) variance of base-experts and 3) difference of client and impostor means. To achieve lower EER, smaller correlation and average variance of base-experts, and larger mean difference are desirable. Furthermore, analysing any of these factors independently, e.g. focusing on correlation alone, could be miss-leading. Experimental results on the BANCA and XM2VTS multi-modal databases and NIST 2001 speaker verification database confirm our findings using VR-EER analysis. Furthermore, F-ratio is shown to be a valid criterion in place of EER as an evaluation criterion. We analysed four commonly

### Revisiting Doddington’s Zoo: A Systematic Method to Assess User-dependent

"... A systematic analysis of user-dependent performance variability in the context of automatic speaker verification was first studied by Doddington et al(1998). Different cate-gories of users were identified and labeled as sheep, goats, lambs and wolves. While this categorization is significant, it doe ..."

Abstract
- Add to MetaCart

(Show Context)
A systematic analysis of user-dependent performance variability in the context of automatic speaker verification was first studied by Doddington et al(1998). Different cate-gories of users were identified and labeled as sheep, goats, lambs and wolves. While this categorization is significant, it does not provide a criterion to rank the users in a database based on their variability in performance. In this work we design and evaluate a user-dependent performance crite-rion that requires only a limited number of client (i.e., gen-uine) training scores. We then extend such a study to for-mulate a user-specific score normalization scheme (a vari-ant of the classical F-norm) and show that user-dependent variabilities can be reduced by employing such a scheme. The results of 13 experiments confirm the efficacy of the pro-posed scheme. 1

### Face Recognition Systems under Spoofing Attack

"... Abstract—Face recognition system is one of the most successful application of computer vision, which has been deployed largely in recent years. In last decade, several algorithms for face recognition have been proposed in the literature. To increase reliability of face recognition systems, the syste ..."

Abstract
- Add to MetaCart

(Show Context)
Abstract—Face recognition system is one of the most successful application of computer vision, which has been deployed largely in recent years. In last decade, several algorithms for face recognition have been proposed in the literature. To increase reliability of face recognition systems, the systems must be able to differentiate between real genuine faces and fake faces (spoofed faces). In this paper, we investigate the robustness under spoofing attack of the well-known face

### S E a R C H

, 2004

"... Combining multiple information sources such as streams (with di#erent features) and multi modal data has shown to be a very promising trend, both in experiments and to some extend in real-life biometric authentication applications. However, combining too many biometric systems (base-experts) will al ..."

Abstract
- Add to MetaCart

Combining multiple information sources such as streams (with di#erent features) and multi modal data has shown to be a very promising trend, both in experiments and to some extend in real-life biometric authentication applications. However, combining too many biometric systems (base-experts) will also increase both hardware and computation costs. Conventional way to selecting a subset of optimal base-experts out of N is to carry out the experiments explicitly. There are 2 1 possible combinations. In this paper, we propose an analytical solution to this task using weighted sum fusion on normalised scores (zero-mean and unit variance). The algorithm depends only on how accurately one can estimate the covariance matrix of the actual test data. The proposed algorithm has a complexcity that is additive between the number of examples and the number of possible combinations while the conventional approach is multiplicative between these two terms. Hence, our approach is more e#cient. It was tested on the BANCA multi-modal database. Experimental results showed that such an algorithm is a viable solution. 1

### Improving Single Modal and Multimodal

, 2004

"... Abstract. This study investigates a new client-dependent normalisation to improve a single biometric authentication system, as well as its effects on fusion. There exists two families of client-dependent normalisation techniques, often applied to speaker authentication. They are client-dependent sco ..."

Abstract
- Add to MetaCart

(Show Context)
Abstract. This study investigates a new client-dependent normalisation to improve a single biometric authentication system, as well as its effects on fusion. There exists two families of client-dependent normalisation techniques, often applied to speaker authentication. They are client-dependent score and threshold normalisation techniques. Examples of the former family of techniques are Z-Norm, D-Norm and T-Norm. There is also a vast amount of literature on the latter family of techniques. Both families are surveyed in this study. Furthermore, we also provide a link between these two families of techniques and show that one is a dual representation of the other. These techniques are intended to adjust the variation across different client models. We propose “F-ratio ” normalisation, or F-Norm, applied to face and speaker authentication systems in two contexts: single modal and fusion of multi-modal biometerics. This normalisation requires that only as few as two client-dependent accesses are available (the more the better). Different from previous normalisation techniques, F-Norm considers the client and impostor distributions simultaneously. We show that F-ratio is a natural choice because it is directly associated to Equal Error Rate. It has the effect of centering the client and impostor distributions such that a global threshold can be easily found. Another difference is that F-Norm actually “interpolates”

### Towards Explaining the Success (Or Failure) of Fusion in Biometric Authentication

, 2005

"... submitted for publication Abstract. Combining multiple information sources, typically from several data streams is a very promising approach, both in experiments and to some extents in various real-life applications. A system that uses more than one behavioural and physiological characteristics to v ..."

Abstract
- Add to MetaCart

(Show Context)
submitted for publication Abstract. Combining multiple information sources, typically from several data streams is a very promising approach, both in experiments and to some extents in various real-life applications. A system that uses more than one behavioural and physiological characteristics to verify whether a person is who he/she claims to be is called a multimodal biometric authentication system. Due to lack of large true multimodal biometric datasets, the biometric trait of a user from a database is often combined with another different biometric trait of yet another user, thus creating a so-called a chimeric user. In the literature, this practice is justified based on the fact that the underlying biometric traits to be combined are assumed to be independent of each other given the user. To the best of our knowledge, there is no literature that approves or disapproves such practice. We study this topic from two aspects: 1) by clarifying the mentioned independence assumption and 2) by constructing a pool of chimeric users from a pool of true modality matched users (or simply “true users”) taken from a bimodal database, such that the performance variability due to chimeric user can be compared with that due to true users. The experimental results suggest that for a large proportion of the experiments, such practice is indeed questionable. Biometric authentication is a

### FÉDÉRALE DE LAUSANNE

"... et de nationalité malaisienne acceptée sur proposition du jury: Prof. J.R. Mosig, président du jury Prof. H. Bourlard, Dr. S. Bengio, directeurs de thèse ..."

Abstract
- Add to MetaCart

et de nationalité malaisienne acceptée sur proposition du jury: Prof. J.R. Mosig, président du jury Prof. H. Bourlard, Dr. S. Bengio, directeurs de thèse