Raymond N.J. Veldhuis

A high quality finger vascular pattern dataset collected using a custom designed capturing device

The number of finger vascular pattern datasets available for the research community is scarce, therefore a new finger vascular pattern dataset containing 1440 images is prsented. This dataset is unique in its kind as the images are of high resolution and have a known pixel density. Furthermore this is the first dataset which contains the age, gender and handedness of the participating volunteers as meta data. The images have been captured using a custom designed capturing device. The various aspects of designing this capturing device are addressed in this paper as well. To confirm whether this new dataset is in fact an important contribution some performance figures in terms of EER of several published state-of-the-art algorithms using this new dataset and an existing dataset from the Peking University are presented. Using this new dataset EERs down to 0.4% have been achieved.

A Bayesian model for predicting face recognition performance using image quality

Quality of a pair of facial images is a strong indicator of the uncertainty in decision about identity based on that image pair. In this paper, we describe a Bayesian approach to model the relation between image quality (like pose, illumination, noise, sharpness, etc) and corresponding face recognition performance. Experiment results based on the MultiPIE data set show that our model can accurately aggregate verification samples into groups for which the verification performance varies fairly consistently. Our model does not require similarity scores and can predict face recognition performance using only image quality information. Such a model has many applications. As an illustrative application, we show improved verification performance when the decision threshold automatically adapts according to the quality of facial images.

Effect of calibration data on forensic likelihood ratio from a face recognition system

A biometric system used for forensic evaluation requires a conversion of the score to a likelihood ratio. A likelihood ratio can be computed as the ratio of the probability of a score given the prosecution hypothesis is true and the probability of a score given the defense hypothesis is true. In this paper we study two different approaches of a forensic likelihood ratio computation in the context of forensic face recognition. These approaches differ in the databases they use to obtain the score distribution under the prosecution and the defense hypothesis and therefore consider slightly different interpretation of these hypotheses. The goal of this study is to quantify the effect of these approaches on the resultant likelihood ratio in the context of evidence evaluation from a face recognition system. A state-of-the art commercial face recognition system is employed for facial images comparison and computation of scores. A simple forensic case is simulated by randomly selecting a small subset from the FRGC database. Images in this subset are used to estimate the score distribution under the prosecution and the defense hypothesis and the effect of different approaches of a likelihood ratio computation is demonstrated and explained. It is observed that there is a significant variation in the resultant likelihood ratios given the databases which are used to model the prosecution and defense hypothesis are varied.

A Study of Identification Performance of Facial Regions from CCTV Images

This paper focuses on automatic face identification for forensic applications. Forensic examiners compare different parts of the face image obtained from a closed-circuit television (CCTV) image with a database of mug shots or good quality image(s) taken from the suspect. In this work we study and compare the discriminative capabilities of different facial regions (also referred to as facial features) such as eye, eyebrow, mouth, etc. It is useful because it can statistically support the current practice of forensic facial comparison. It is also of interest to biometrics as a more robust general-purpose face recognition system can be built by fusing the similarity scores obtained from the comparison of different individual parts of the face. For experiments with automatic systems, we simulate a very challenging recognition scenario by using a database of 130 subjects each having only one gallery image. Gallery images are frontal mug shots while probe set consist of low quality CCTV camera images. Face images in gallery and probe sets are first segmented using eye locations and recognition experiments are performed for the different face regions considered. We also study and evaluate an improved recognition approach based on AdaBoost algorithm with Linear Discriminant Analysis (LDA) as a week learner and compare its performance with the baseline Eigenface method for automatic facial feature recognition.

Impact of eye detection error on face recognition performance

The locations of the eyes are the most commonly used features to perform face normalisation (i.e. alignment of facial features), which is an essential preprocessing stage of many face recognition systems. In this study, the authors study the sensitivity of open source implementations of five face recognition algorithms to misalignment caused by eye localisation errors. They investigate the ambiguity in the location of the eyes by comparing the difference between two independent manual eye annotations. They also study the error characteristics of automatic eye detectors present in two commercial face recognition systems. Furthermore, they explore the impact of using different eye detectors for training/enrolment and query phases of a face recognition system. These experiments provide an insight into the influence of eye localisation errors on the performance of face recognition systems and recommend a strategy for the design of training and test sets of a face recognition algorithm.

Evaluation of AFIS-Ranked Latent Fingerprint Matched Templates

The methodology currently practiced in latent print examination (known as ACE-V) yields only a decision as result, namely individualization, exclusion or inconclusive. From such a decision, it is not possible to express the strength of opinion of a forensic examiner quantitatively with a scientific basis to the criminal justice system. In this paper, we propose a framework to generate a score from the matched template generated by the forensic examiner. Such a score can be viewed as a measure of confidence of a forensic examiner quantitatively, which in turn can be used in statistics-based evidence evaluation framework, for e.g, likelihood ratio. Together with the description and evaluation of new realistic forensic case driven score computation, we also exploit the developed experimental framework to understand more about matched templates in forensic fingerprint databases.

Identification Performance of Evidential Value Estimation for Fingermarks

Law enforcement agencies around the world use biometrics and fingerprints to solve and fight crime. Forensic experts are needed to record fingermarks at crime scenes and to ensure those captured are of evidential value. This process needs to be automated and streamlined as much as possible to improve efficiency and reduce workload. It has previously been demonstrated that is possible to estimate a fingermarks evidential value automatically for image captures taken with a mobile phone or other devices, such as a scanner or a high-quality camera. Here we study the relationship between a fingermark being of evidential value and its correct and certain identification and if it is possible to achieve identification despite the mark not having sufficient evidential value. Subsequently, we also investigate the influence the capture device used makes and if a mobile phone is a considerable option. Our results show that automatic identification is possible for 126 of the 1,428 fingermarks captured by a mobile phone, of which 116 were marked as having evidential value by experts and 123 by an automated algorithm.

Sampling variability in forensic likelihood-ratio computation: A simulation study

Recently, in the forensic biometric community, there is a growing interest to compute a metric called likelihood-ratio when a pair of biometric specimens is compared using a biometric recognition system. Generally, a biometric recognition system outputs a score and therefore a likelihood-ratio computation method is used to convert the score to a likelihood-ratio. The likelihood-ratio is the probability of the score given the hypothesis of the prosecution, Hp (the two biometric specimens arose from a same source), divided by the probability of the score given the hypothesis of the defense, Hd (the two biometric specimens arose from different sources). Given a set of training scores under Hp and a set of training scores under Hd, several methods exist to convert a score to a likelihood-ratio. In this work, we focus on the issue of sampling variability in the training sets and carry out a detailed empirical study to quantify its effect on commonly proposed likelihood-ratio computation methods. We study the effect of the sampling variability varying: 1) the shapes of the probability density functions which model the distributions of scores in the two training sets; 2) the sizes of the training sets and 3) the score for which a likelihood-ratio is computed. For this purpose, we introduce a simulation framework which can be used to study several properties of a likelihood-ratio computation method and to quantify the effect of sampling variability in the likelihood-ratio computation. It is empirically shown that the sampling variability can be considerable, particularly when the training sets are small. Furthermore, a given method of likelihood-ratio computation can behave very differently for different shapes of the probability density functions of the scores in the training sets and different scores for which likelihood-ratios are computed.

Likelihood ratio based mixed resolution facial comparison

In this paper, we propose a novel method for low-resolution face recognition. It is especially useful for a common situation in forensic search where faces of low resolution, e.g. on surveillance footage or in a crowd, must be compared to a high-resolution reference. This method is based on the likelihood ratio of a pair of mixed-resolution input images. The effectiveness of our method is tested on the SCface database which contains face images taken by surveillance cameras. The results show that our method outperforms recently published state-of-the-art.

Regional fusion for high-resolution palmprint recognition using spectral minutiae representation

The spectral minutiae representation (SMC) has been recently proposed as a novel method to minutiae-based fingerprint recognition, which is invariant to minutiae translation and rotation and presents low computational complexity. As high-resolution palmprint recognition is also mainly based on minutiae sets, SMC has been applied to palmprints and used in full-to-full palmprint matching. However, the performance of that approach was still limited. As one of the main reasons for this is the much bigger size of a palmprint compared with a fingerprint, the authors propose a division of the palmprint into smaller regions. Then, to further improve the performance of spectral minutiae-based palmprint matching, in this work the authors present anatomically inspired regional fusion while using SMC for palmprints. Firstly, the authors consider three regions of the palm, namely interdigital, thenar and hypothenar, which have inspiration in anatomic cues. Then, the authors apply SMC to region-to-region palmprint comparison and study regional discriminability when using the method. After that, the authors implement regional fusion at score level by combining the scores of different regional comparisons in the palm with two fusion methods, that is, sum rule and logistic regression. The authors evaluate region-to-region comparison and regional fusion based on spectral minutiae matching on a public high-resolution palmprint database, THUPALMLAB. Both manual segmentation and automatic segmentation are performed to obtain the three palm regions for each palm. Essentially using the complex SMC, the authors obtain results on region-to-region comparison which show that the hypothenar and interdigital regions outperform the thenar region. More importantly, the authors achieve significant performance improvements by regional fusion using regions segmented both manually and automatically. One main advantage of the approach the authors took is that human examiners can segment the palm into the three regions without prior knowledge of the system, which makes the segmentation process easy to be incorporated in protocols such as in forensic science.