Sarat C. Dass

Likelihood Ratio-Based Biometric Score Fusion

Multibiometric systems fuse information from different sources to compensate for the limitations in performance of individual matchers. We propose a framework for the optimal combination of match scores that is based on the likelihood ratio test. The distributions of genuine and impostor match scores are modeled as finite Gaussian mixture model. The proposed fusion approach is general in its ability to handle 1) discrete values in biometric match score distributions, 2) arbitrary scales and distributions of match scores, 3) correlation between the scores of multiple matchers, and 4) sample quality of multiple biometric sources. Experiments on three multibiometric databases indicate that the proposed fusion framework achieves consistently high performance compared to commonly used score fusion techniques based on score transformation and classification.

Soft Biometric Traits for Personal Recognition Systems

Many existing biometric systems collect ancillary information like gender, age, height, and eye color from the users during enrollment. However, only the primary biometric identifier (fingerprint, face, hand-geometry, etc.) is used for recognition and the ancillary information is rarely utilized. We propose the utilization of “soft” biometric traits like gender, height, weight, age, and ethnicity to complement the identity information provided by the primary biometric identifiers. Although soft biometric characteristics lack the distinctiveness and permanence to identify an individual uniquely and reliably, they provide some evidence about the user identity that could be beneficial. This paper presents a framework for integrating the ancillary information with the output of a primary biometric system. Experiments conducted on a database of 263 users show that the recognition performance of a fingerprint system can be improved significantly (≈ 5%) by using additional user information like gender, ethnicity, and height.

Fingerprint Quality Indices for Predicting Authentication Performance

The performance of an automatic fingerprint authentication system relies heavily on the quality of the captured fingerprint images. In this paper, two new quality indices for fingerprint images are developed. The first index measures the energy concentration in the frequency domain as a global feature. The second index measures the spatial coherence in local regions. We present a novel framework for evaluating and comparing quality indices in terms of their capability of predicting the system performance at three different stages, namely, image enhancement, feature extraction and matching. Experimental results on the IBM-HURSLEY and FVC2002 DB3 databases demonstrate that the global index is better than the local index in the enhancement stage (correlation of 0.70 vs. 0.50) and comparative in the feature extraction stage (correlation of 0.70 vs. 0.71). Both quality indices are effective in predicting the matching performance, and by applying a quality-based weighting scheme in the matching algorithm, the overall matching performance can be improved; a decrease of 1.94% in EER is observed on the FVC2002 DB3 database.

Localized iris image quality using 2-d wavelets

The performance of an iris recognition system can be undermined by poor quality images and result in high false reject rates (FRR) and failure to enroll (FTE) rates. In this paper, a wavelet-based quality measure for iris images is proposed. The merit of the this approach lies in its ability to deliver good spatial adaptivity and determine local quality measures for different regions of an iris image. Our experiments demonstrate that the proposed quality index can reliably predict the matching performance of an iris recognition system. By incorporating local quality measures in the matching algorithm, we also observe a relative matching performance improvement of about 20% and 10% at the equal error rate (EER), respectively, on the CASIA and WVU iris databases.

A principled approach to score level fusion in multimodal biometric systems

A multimodal biometric system integrates information from multiple biometric sources to compensate for the limitations in performance of each individual biometric system. We propose an optimal framework for combining the matching scores from multiple modalities using the likelihood ratio statistic computed using the generalized densities estimated from the genuine and impostor matching scores. The motivation for using generalized densities is that some parts of the score distributions can be discrete in nature; thus, estimating the distribution using continuous densities may be inappropriate. We present two approaches for combining evidence based on generalized densities: (i) the product rule, which assumes independence between the individual modalities, and (ii) copula models, which consider the dependence between the matching scores of multiple modalities. Experiments on the MSU and NIST multimodal databases show that both fusion rules achieve consistently high performance without adjusting for optimal weights for fusion and score normalization on a case-by-case basis.

Quality-based Score Level Fusion in Multibiometric Systems

The quality of biometric samples has a significant impact on the accuracy of a matcher. Poor quality biometric samples often lead to incorrect matching results because the features extracted from these samples are not reliable. Therefore, dynamically assigning weights to the outputs of individual matchers based on the quality of the samples presented at the input of the matchers can improve the overall recognition performance of a multibiometric system. We propose a likelihood ratio-based fusion scheme that takes into account the quality of the biometric samples while combining the match scores provided by the matchers. Instead of estimating the quality of the template and query images individually, we estimate a single quality metric for each template-query pair based on the local image quality measures. Experiments on a database of 320 users with iris and fingerprint modalities demonstrate the advantages of utilizing the quality information in multibiometric systems

A Total Variation-Based Algorithm for Pixel-Level Image Fusion

In this paper, a total variation (TV) based approach is proposed for pixel-level fusion to fuse images acquired using multiple sensors. In this approach, fusion is posed as an inverse problem and a locally affine model is used as the forward model. A TV semi-norm based approach in conjunction with principal component analysis is used iteratively to estimate the fused image. The feasibility of the proposed algorithm is demonstrated on images from computed tomography (CT) and magnetic resonance imaging (MRI) as well as visible-band and infrared sensors. The results clearly indicate the feasibility of the proposed approach.

Markov random field models for directional field and singularity extraction in fingerprint images

A Bayesian formulation is proposed for reliable and robust extraction of the directional field in fingerprint images using a class of spatially smooth priors. The spatial smoothness allows for robust directional field estimation in the presence of moderate noise levels. Parametric template models are suggested as candidate singularity models for singularity detection. The parametric models enable joint extraction of the directional field and the singularities in fingerprint impressions by dynamic updating of feature information. This allows for the detection of singularities that may have previously been missed, as well as better aligning the directional field around detected singularities. A criteria is presented for selecting an optimal block size to reduce the number of spurious singularity detections. The best rates of spurious detection and missed singularities given by the algorithm are 4.9% and 7.1%, respectively, based on the NIST 4 database.

Statistical Models for Assessing the Individuality of Fingerprints

Following the Daubert ruling in 1993, forensic evidence based on fingerprints was first challenged in the 1999 case of the U.S. versus Byron C. Mitchell and, subsequently, in 20 other cases involving fingerprint evidence. The main concern with the admissibility of fingerprint evidence is the problem of individualization, namely, that the fundamental premise for asserting the uniqueness of fingerprints has not been objectively tested and matching error rates are unknown. In order to assess the error rates, we require quantifying the variability of fingerprint features, namely, minutiae in the target population. A family of finite mixture models has been developed in this paper to represent the distribution of minutiae in fingerprint images, including minutiae clustering tendencies and dependencies in different regions of the fingerprint image domain. A mathematical model that computes the probability of a random correspondence (PRC) is derived based on the mixture models. A PRC of 2.25 times10-6 corresponding to 12 minutiae matches was computed for the NIST4 Special Database, when the numbers of query and template minutiae both equal 46. This is also the estimate of the PRC for a target population with a similar composition as that of NIST4.

Validating a Biometric Authentication System: Sample Size Requirements

Authentication systems based on biometric features (e.g., fingerprint impressions, iris scans, human face images, etc.) are increasingly gaining widespread use and popularity. Often, vendors and owners of these commercial biometric systems claim impressive performance that is estimated based on some proprietary data. In such situations, there is a need to independently validate the claimed performance levels. System performance is typically evaluated by collecting biometric templates from n different subjects, and for convenience, acquiring multiple instances of the biometric for each of the n subjects. Very little work has been done in 1) constructing confidence regions based on the ROC curve for validating the claimed performance levels and 2) determining the required number of biometric samples needed to establish confidence regions of prespecified width for the ROC curve. To simplify the analysis that addresses these two problems, several previous studies have assumed that multiple acquisitions of the biometric entity are statistically independent. This assumption is too restrictive and is generally not valid. We have developed a validation technique based on multivariate copula models for correlated biometric acquisitions. Based on the same model, we also determine the minimum number of samples required to achieve confidence bands of desired width for the ROC curve. We illustrate the estimation of the confidence bands as well as the required number of biometric samples using a fingerprint matching system that is applied on samples collected from a small population