Sergey Tulyakov

Review of Classifier Combination Methods

Classifier combination methods have proved to be an effective tool to increase the performance of pattern recognition applications. In this chapter we review and categorize major advancements in this field. Despite a significant number of publications describing successful classifier combination implementations, the theoretical basis is still missing and achieved improvements are inconsistent. By introducing different categories of classifier combinations in this review we attempt to put forward more specific directions for future theoretical research. We also introduce a retraining effect and effects of locality based training as important properties of classifier combinations. Such effects have significant influence on the performance of combinations, and their study is necessary for complete theoretical understanding of combination algorithms.

Symmetric hash functions for secure fingerprint biometric systems

Securing biometrics databases from being compromised is an important research challenge that must be overcome in order to support widespread use of biometrics based authentication. In this paper we present a novel method for securing fingerprints by hashing the fingerprint minutia and performing matching in the hash space. Our approach uses a family of symmetric hash functions and does not depend on the location of the (usually unstable) singular points (core and delta) as is the case with other methods described in the literature. It also does not assume a pre-alignment between the test and the stored fingerprint templates. We argue that these assumptions, which are often made, are unrealistic given that fingerprints are very often only partially captured by the commercially available sensors. The Equal Error Rate (EER) achieved by our system is 3%. We also present the performance analysis of a hybrid system that has an EER of 1.96% which reflects almost no drop in performance when compared to straight matching with no security enhancements. The hybrid system involves matching using our secure algorithm but the final scoring reverts to that used by a straight matching system.

Symmetric hash functions for fingerprint minutiae

The possibility that a biometric database is compromised is one of the main concerns in implementing biometric identification systems. The compromise of a biometric renders it permanently useless. In this paper we present a method of hashing fingerprint minutia information and performing fingerprint identification in a new space. Only hashed data is transmitted and stored in the server database, and it is not possible to restore fingerprint minutia locations using hashed data. We also present a performance analysis of the proposed algorithm.

Robust point-based feature fingerprint segmentation algorithm

A critical step in automatic fingerprint recognition is the accurate segmentation of fingerprint images. The objective of fingerprint segmentation is to decide which part of the images belongs to the foreground containing features for recognition and identification, and which part to the background with the noisy area around the boundary of the image. Unsupervised algorithms extract blockwise features. Supervised method usually first extracts point features like coherence, average gray level, variance and Gabor response, then a Fisher linear classifier is chosen for classification. This method provides accurate results, but its computational complexity is higher than most of unsupervised methods. This paper proposes using Harris corner point features to discriminate foreground and background. Shifting a window in any direction around the corner should give a large change in intensity. We observed that the strength of Harris point in the foreground area is much higher than that of Harris point in background area. The underlying mechanism for this segmentation method is that boundary ridge endings are inherently stronger Harris corner points. Some Harris points in noisy blobs might have higher strength, but it can be filtered as outliers using corresponding Gabor response. The experimental results proved the efficiency and accuracy of new method are markedly higher than those of previously described methods.

Probabilistic model for segmentation based word recognition with lexicon

We describe the construction of a model for off-line word recognizers based on over-segmentation of the input image and recognition of segment combinations as characters in a given lexicon word. One such recognizer, the Word Model Recognizer (WMR), is used extensively. Based on the proposed model it was possible to improve the performance of WMR.

Comparison of combination methods utilizing T-normalization and second best score model

The combination of biometric matching scores can be enhanced by taking into account the matching scores related to all enrolled persons in addition to traditional combinations utilizing only matching scores related to a single person. Identification models take into account the dependence between matching scores assigned to different persons and can be used for such enhancement. In this paper we compare the use of two such models - T-normalization and second best score model. The comparison is performed using two combination algorithms - likelihood ratio and multilayer perceptron. The results show, that while second best score model delivers better performance improvement than T-normalization, two models are complementary to each other and can be used together for further improvements.

Combination of Symmetric Hash Functions for Secure Fingerprint Matching

Fingerprint based secure biometric authentication systems have received considerable research attention lately, where the major goal is to provide an anonymous, multipliable and easily revocable methodology for fingerprint verification. In our previous work, we have shown that symmetric hash functions are very effective in providing such secure fingerprint representation and matching since they are independent of order of minutiae triplets as well as location of singular points (e.g. core and delta). In this paper, we extend our prior work by generating a combination of symmetric hash functions, which increases the security of fingerprint matching by an exponential factor. Firstly, we extract kplets from each fingerprint image and generate a unique key for combining multiple hash functions up to an order of (k-1). Each of these keys is generated using the features extracted from minutiae k-plets such as bin index of smallest angles in each k-plet. This combination provides us an extra security in the face of brute force attacks, where the compromise of few hash functions as well do not compromise the overall matching. Our experimental results suggest that the EER obtained using the combination of hash functions (4.98%) is comparable with the baseline system (3.0%), with the added advantage of being more secure.

Facial Expression Biometrics Using Tracker Displacement Features

In this paper we investigate a possibility of using the face expression information for person biometrics. The idea of this research is that persons emotional face expressions are repeatable, and face expression features can be used for person identification. In order to avoid using person specific geometric or textural features traditionally used in face biometrics, we restrict ourselves to the tracker displacement features only. In contrast to previous research in facial expression biometrics, we extract features only from the pair of face images, neutral and the apex of emotion expression, instead of using the sequence of images from the video. The experiments, performed on two facial expression databases, confirm that proposed features can indeed be used for biometrics purposes.

Robust Real-Time Extreme Head Pose Estimation

This paper proposes a new framework for head pose estimation under extreme pose variations. By augmenting the precision of a template matching based tracking module with the ability to recover offered by a frame-by-frame head pose estimator, we are able to address pose ranges for which face features are no longer visible, while maintaining state-of-the-art performance. Experimental results obtained on a newly acquired 3D extreme head pose dataset support the proposed method and open new perspectives in approaching real-life unconstrained scenarios.

A Framework for Efficient Fingerprint Identification Using a Minutiae Tree

Given the existence of large fingerprint databases, including distributed systems, the development of algorithms for performing fast searches in them has become the important topic for biometric researchers. In this paper, we propose a new indexing method for fingerprint templates consisting of a set of minutia points. In contrast to previously presented methods, our algorithm is tree-based and well addresses the efficiency needs of complex (possibly distributed) systems. One large index tree is constructed and the enrolled templates are represented by the leaves of the tree. The branches in the index tree correspond to different local configurations of minutia points. Searching the index tree entails extracting local minutia neighborhoods of the test fingerprint and matching them against tree nodes. Therefore, the search time does not depend on the number of enrolled fingerprint templates, but only on the index tree configuration. This framework can be adapted for different tree-building parameters (feature sets, indexing levels, bin boundaries) according to user requirements and different enrollment and searching techniques can be applied to improve accuracy. We conduct a number of the experiments on Fingerprint Verification Competition databases, as well as the databases of synthetically generated fingerprint templates. The experiments confirm the ability of the proposed algorithm to find correct matches in the database and the minimum search time requirements.