Berkay Topcu

An acoustic based sniper detection system prototype

In this work, an acoustic-based sniper detection system prototype geometry and its operational principals are presented from the signal processing perspective. The prototype consists of a microphone network positioned in a specific geometric structure in the three dimensional space. The system depends on estimating the delay of arrival of sound waves reaching the microphones and using the delay information in order to calculate the direction of the sound source. The time difference of arrival problem is solved by using the generalized cross correlation approach. The estimated delay is then transformed into angles using the far field approximation. The found angle is the angle between the sound source and the microphone pair axis. Using the angles calculated for different microphone pairs, the direction is reported with azimuth and elevation. The study reports the simulation results and laboratory experiments.

How to assess privacy preservation capability of biohashing methods?: Privacy metrics

In this paper, we evaluate privacy preservation capability of biometric hashing methods. Although there are some work on privacy evaluation of biometric template protection methods in the literature, they fail to cover all biometric template protection methods. To the best of our knowledge, there is no work on privacy metrics and assessment for biometric hashing methods. We use several metrics under different threat scenarios to assess privacy protection level of biometric hashing methods in this work. The simulation results demonstrate that biometric hash vectors may leak private information especially under advanced threat scenarios.

Fingerprint phalanx-based score fusion

Traditional fingerprint-based authentication systems work on images obtained (e.g., via optical, thermal sensors) from the first phalanges of human fingers. Typical minutia-based systems extract the relevant feature vectors and these feature vectors are utilized by matchers. In this study, we report minutiae matching performance figures, originating from adaptively fusing second and third phalanx images of a mid-size fingerprint database that we collected in our laboratory. In a prior publication, we analyzed the individual phalanges in terms of matching accuracy, image quality, and feature statistics. As a continuation of that prior work, the algorithm developed here (using image quality in score weighting) leads to experimentally verified performance improvement. Utilizing confidence intervals for measuring statistical significance, the performance figures indicate that for cases where first phalanx images are not usable (e.g. due to missing digits, low image quality due to manual labor), proposed fusion algorithm provides an alternative authentication mechanism. Further, when all three phalanges are used, the authentication performance increases with respect to the first-phalanx-only scenario.

Correlation-based patch localization for face recognition

With patch-based approaches, it is aimed to tackle the factors such as illumination, pose changes and partial occlusions that are faced in real world applications and complicates the face recognition problem. For patch-based face recognition systems to work robustly, patch locations should correspond to similar image content. In this paper, we propose two patch localization schemes for patch-based face recognition in order to make patch locations to correspond to same area in all of the face images and the image contents of the patches as close as possible. Our experimental results show that with either of the localization schemes, higher recognition results are obtained especially on the partially occluded face images with scarves or sunglasses.

Unpredictability assessment of biometric hashing under naive and advanced threat conditions

Recent years have witnessed the use of biometric recognition systems in increasing number of applications with the number of users growing at a steady pace. However, security and privacy problems have arisen from this upsurge of interest to biometric systems. Template protection methods solve such security and privacy problems where unpredictability is a crucial goal. Here, we study the unpredictability of biohashing (a transformation-based template protection method) using entropy as a measure. Our novel work outlines a systematic approach for theoretical evaluation of biohashes using estimated entropy which is based on degree of freedom of Binomial distribution. Our experiments demonstrate that biohash unpredictability varies in different threat models where the entropy of a biohash is almost equal to its bit length under the naive scenario and is significantly low in the advanced scenario, implying that the amount of information kept hidden in a biohash is more likely to be predicted.

Turkish Normalization Lexicon for Social Media

Social media has its own evergrowing language and distinct characteristics. Although social media is shown to be of great utility to research studies, varying quality of written texts degrades the performance of existing NLP tools. Normalization of texts, transforming from informal to well-written texts, appears to be a reasonable preprocessing step to adapt tools trained on different domains to social media. In this study, we compile the first Turkish normalization lexicon that sheds light to the kinds of observed lexical variations in social media texts. A graphical representation acquired from a text corpus is used to model contextual similarities between normalization equivalences and the lexicon is automatically generated by performing random walks on this graph. The underlying framework not only enables different lexicons to be generated from the same corpus but also produces lexicons that are tuned to specific genres. Evaluation studies demonstrated the effectiveness of induced lexicon in normalizing Turkish texts.

GMM-SVM Fingerprint Verification Based on Minutiae Only

Most fingerprint recognition systems use minutiae information, which is an unordered collection of minutiae locations and orientations. Template protection algorithms such as fuzzy commitment and other modern cryptographic alternatives based on homomorphic encryption require a fixed size binary template. However, such a template is not directly applicable to fingerprint minutiae representation which by its nature is of variable size. In this study, we introduce a novel method to represent a minutiae set with a rotation invariant fixed-length vector. We represent each minutia according to its geometric relation with neighbors and use Gaussian mixture model (GMM) to model its feature distribution. A two-class linear SVM is used to create a model template for the enrollment fingerprint sample, which discriminates impressions of the same finger from other fingers. We evaluated the verification performance of our method on the FVC2002DB1 database.

Biohashing with Local Zernike Moments for face verification

Local Zernike moments (LZM) is a recently proposed image representation scheme that is shown to be successful for face representation. In this study, a face verification system which depends on biometric hashing scheme, that uses LZM features extracted from face images is proposed. With the proposed system, security and user privacy is ensured. Verification performance of the system, in which user specific secret keys are used for biometric hashing, is evaluated in two different scenarios, on the BioSecure face database. In the first scenario, biometric hashing is realized using each users secret key and %0 equal error is obtained. In the second scenario, in which the secret key of a user is stolen by an adversary, biometric hashes are created using the stolen key and any biometric sample. In this case, the equal error rate increases to %8, 26, which is comparable to equal error rate of %6, 81, where only LZM feature vectors are used for verification.

TÜBİTAK-BİLGEM Shot Estimation System (SES)

In this work, design and experimental studies related to TUBITAK-BILGEM Shot Estimation System (AKS in Turkish) will be discussed. AKS is composed of three parts which are, a microphone array that has a specific structural design, an electronic unit that enables synchronous recording of the microphone signals and a graphical user interface that prompts the output of the system. Basic goal is to detect the position of a pre-defined sound source in terms of cartesian coordinates with respect to the original position of the system. First, the position of the sound source is detected in the cartesian coordinates by using only the time difference of arrival information. The time difference of arrival problem is solved using the generalized cross correlation function. In order to compensate for the false alarms that is very common in these systems, an energy based and a Mel Frequency Cepstrum Coefficients based two stage classifier is used. System is tested with shots from four different hand held rifles, from two different distances for each 10 degree angles spreading over 180 degrees. The average directional error for 100m shots is 2,69 degrees while the error declines to 1,51 degrees for 200m shots. The precision of the system is %84,6 while the recall is %96,2.