home  

publications

graduate students                

bibtex file

 

 

This page lists my recent research work. For more details, you can contact me and check the publications  as it may be more often updated.

 

Announcement:

 

I am no longer looking for students, except possibly for 1-2 new students, because I already have too many.

 

The applications should be done directly to Sabanci University during the graduate student admission periods (April-May period).

Tuition waiver to cover University expenses and stipend to cover living expenses will be given to successful applicants.

On-campus housing is also available.

 

There is usually room to work for ENS491-492 students or workstudy students, in one of those projects.

If interested, please contact me with your details (interest, GPA, courses taken, …).

 

 

 

 

 

 

 

Biometrics

 

 

Biometric technologies are used to recognize a person from his/her biological traits such as fingerprint, iris, retina, face, handshape, signature, keystroke dynamics,... It is increasingly used in identity verification, while also drawing concerns about a loss of privacy. We have developed verification systems for online and offline signatures, fingerprints, and voice.

 

A newly studied aspect of biometrics is the issue of privacy: what happens if someone steals your fingerprint? You cannot replace nor revoke it! We study methods to prevent privacy loss that accompany the benefits of biometrics (e.g. cancelable biometrics, combined biometrics, fuzzy vault etc.). Biometric cryptosystems, systems combining biometrics and cryptography, tries to achieve cryptographic systems based on biometrics, and also address the privacy problem as well.

 

We have suggested a novel approach of combining two fingerprints (and later a fingerprint with voice) to create a multi-biometric template to improve BOTH security and privacy of the user (by protecting the individual biometric).

  • Yanıkoğlu, Berrin and Kholmatov, Alisher Anatolyevich, "Combining multiple biometrics to protect privacy", Proceedings of ICPR-BCTP Workshop, 2004.
  • Eren Camlikaya, Alisher Kholmatov and Berrin Yanikoglu, "Multi-biometric Templates Using Fingerprint and Voice", Biometric technology for human identification, Vol. 6944, no5, pp: 1-9, Orlando FL, 2008.

 

 The online signature verification system we developed has won the first place at The First International Signature Verification Competition (SVC 2004) organized in conjunction with the First International Conference on Biometric Authentication (ICBA 2004)  and we received first place in many subsequent online and offline signature verification competitions.

 

Related papers:

  • Yılmaz, Mustafa Berkay and Yanıkoğlu, Berrin and Tırkaz, Çağlar and Kholmatov, Alisher Anatolyevich, "Offline signature verification using classifier combination of HOG and LBP features", International Joint Conference on Biometrics (IJCB 2011), USA: IEEE (Institute of Electrical and Electronics Engineers), November 2011, 1-7
  • Kholmatov, Alisher Anatolyevich and Yanıkoğlu, Berrin, "SUSIG: an on-line signature database, associated protocols and benchmark results", Pattern Analysis & Applications, Vol.12, No.3, September 2009, 227-236 (SCI)
  • Yanıkoğlu, Berrin and Kholmatov, Alisher, "Online signature verification using fourier descriptors", EURASIP Journal on Advances in Signal Processing 2009 (SCI)
  • Kholmatov, Alisher Anatolyevich and Yanıkoğlu, Berrin, "Realization of correlation attack against fuzzy vault scheme", Security, Forensics, Steganography, and Watermarking of Multimedia Contents X, USA: SPIE (The International Society for Optical Engineering), January 2008
  • Yanıkoğlu, Berrin and Çamlıkaya, Eren and Erdoğan, Hakan, "Saklı modeli kullanarak ses ile metin bağımlı kimlik doğrulama ", Sinyal İşleme ve İletişim Uygulamaları Kurultayı SİU 2007, Eskişehir, Turkey: IEEE, June 2007
  • Kholmatov, Alisher Anatolyevich and Yanıkoğlu, Berrin, "Biometric cryptosystem using online signatures", Lecture Notes in Computer Science, Vol.4263, November 2006, 981-990 (SCI)
  • Yanıkoğlu, Berrin and Levi, Albert and Savaş, Erkay and Kholmatov, Alisher Anatolyevich, "Secret sharing using biometric traits", Biometric Technology For Human Identification, Orlando, FL, USA: SPIE, April 2006.
  • Kholmatov, Alisher Anatolyevich and Yanıkoğlu, Berrin, "Identity authentication using improved online signature verification method", Pattern recognition letters, Vol.26, No.15, November 2005, 2400-2408 (SCI).

·        

Thesis supervised:

 

Projects:

·       TÜBİTAK Project (No: 105 E 165 Duration: 6/2006-6/2008):  “Privacy Protecting Biometric Authentication Systems”.

 

 

 

 

 

Machine Learning

 

We are working on optimization of a classifier combination technique called ECOC (Error Correcting Output Codes). Classifier combinations achieve better generalization errors compared to individual classifiers, due to lower variance.  In ECOC, classifier combination is done such that each class has a codeword indicating the assignment of base classifiers and the input is assigned to the nearest codeword, in an error correcting fashion.

 

Related papers:

  • Zor, Cemre and Windeatt, Terry and Yanıkoğlu, Berrin, "Bias-variance analysis of ECOC and bagging using neural nets", SUEMA 2010 Workshop - Supervised and Unsupervised Ensemble Methods and their Applications, Barcelona, Spain: Pattern Analysis, Statistical Modelling and Computational Learning (PASCAL 2), September 2010, 109-118
  • Zor, Cemre and Yanıkoğlu, Berrin and Windeatt, Terry and Alpaydın, Ethem, "FLIP-ECOC: a greedy optimization of the ECOC matrix", 25th International Symposium on Computer and Information Sciences (ISCIS 2010), Germany: Springer, September 2010

 

 

 

 

 

Handwriting & Sketch Recognition

 

Handwriting recognition is the problem of recognizing a handwritten word; the input may be obtained via scanning (offline recognition) or from a digitizing tablet (online recognition). Whereas recognizing machine print (OCR) is largely solved for clean documents, the problem still remains for noisy, small font documents, and handwritten documents. We also work on the problem of mathematical expression recognition, with the goal of converting handwritten expression into LaTeX and sketch recognition for interpreting sketches drawn on a tablet.

 

Related papers:

  • Çelik, Mehmet and Yanıkoğlu, Berrin, "Probabilistic mathematical formula recognition using a 2D context-free graph grammar", International Conference on Document Analysis and Recognition (ICDAR 2011), USA: IEEE (Institute of Electrical and Electronics Engineers) 2011, 161-166.
  • Tırkaz, Çağlar and Sezgin, T. Metin and Yanıkoğlu, Berrin, "Sketched symbol recognition with few examples using particle filtering", ACM International Symposium on Sketch-Based Interfaces and Modeling, Vancouver, CA: ACM (Association for Computing Machinery), August 2011.
  • Büyükbayrak , Hakan and Yanıkoğlu, Berrin and Erçil, Aytül, "Online handwritten mathematical expression recognition", Lin, Xiaofan and Yanıkoğlu, Berrin (eds.), Document Recognition and Retrieval XIV: IS&T/SPIE International Symposium on Electronic Imaging (SPIE 2007), San Jose, California, USA:, January 2007.
  •  "Turkish handwritten text recognition: A Case of Agglutinative Languages". Proceedings of SPIE, Jan 2003.
  • "Text Detection and Extraction in Outdoor Scenes", Alisher Kholmatov, Aytül Erçil, Berrin Yanikoglu, IEEE SIU Conference, June 12-14, 2002 Pamukkale, Turkey.
  • "Pitch Estimation and Pitch-Based Segmentation for Dot-Matrix Text Recognition", Berrin Yanikoglu, International Journal on Document Analysis and Recognition, 3, 1, 2000.
  • "Segmentation of Off-line Cursive Handwriting Using Linear Programming", Berrin Yanikoglu and Peter Sandon, "Patter Recognition" 31, 12 1998.

 

Thesis supervised (w/ Prof. Erçil):

 

 

Papers in Turkish:

  • Vural, Esra and Erdoğan, Hakan and Oflazer, Kemal and Yanıkoğlu, Berrin, "Türkçe için tablet PC ortamında çevrimiçi yazı tanıma", SIU 2004, IEEE, May 2004, 607-610
  • Vural, Esra and Erdoğan, Hakan and Oflazer, Kemal and Yanıkoğlu, Berrin, "Türkçe için tablet PC ortamında çevrimiçi yazı tanıma sistemi = An online handwriting recognition system for Turkish", IEEE 12th Signal Processing and Communications Applications Conference, 2004, USA: IEEE (Institute of Electrical and Electronics Engineers) 2004, 607-610

 

Projects:

 

  • TÜBİTAK Project (No: 107E271 Duration: 12/2008-11/2010):  “Handwritten Mathematical Expression Recognition”.

·       TÜBİTAK Project (No: 101E012 Duration: 9/2001-9/2003):  “Document segmentation and Recognition for Turkish”: In this project, we developed image processing and document recognition algorithms suitable for Turkish, since the agglutinative nature of Turkish morphology prevents the use of previously developed OCR systems.

 

 

 

 

 

 

Image Retrieval

 

We work on plant image retrieval problem, with the goal of locating information about plants using images of a plant.

 

 

Related papers:

·       Kebapcı, Hanife and Yanıkoğlu, Berrin and Ünal, Gözde, "Plant image retrieval using color, shape and texture features", Computer Journal (SI), Vol.54, No.9, September 2011, 1475-1490 (SCI).

 

·       Yanıkoğlu, Berrin and Aptoula, Erchan and Tırkaz, Çağlar, "Sabanci-Okan system at ImageClef 2011: plant identication task", Petras, Vivien and Forner, Pamela and Clough, Paul D. (eds.), CLEF 2011 - Conference on Multilingual and Multimodal Information Access Evaluation, Amsterdam, Netherlands: CLEF (Cross-Language Evaluation Forum) 2011.

 

·       We have been participating in the Plant Identification campaign organized by ImageCLEF/LifeCLEF. Hence we have papers in subsequent workshops as well.

 

Thesis supervised:

  • Kebapcı, Hanife (2009) Content Based Image Retrieval for Identification of House Plants Using Color, Texture and Shape Features [MS Thesis ]
  • S. Tolga Yıldıran (expected graduation, Fall 2015)
  • Mostafa Mehdipour Ghazi (expected graduation, Summer 2015)

 

 

 

 

 

Document Analysis and Recognition

 

Document analysis refers to understanding the logical (headlines, footnotes,...) and physical structure (separate columns, halftones,...) of a document, so that a scanned page can be reconstructed in digital form, exactly matching the original image. The main task is to identify the separate physical regions (images from text, columns boundaries), while another objective is to understand the purpose of a particular region (e.g. footnote), so as to put it in its proper context.

 

Document segmentation aims to segment a document into its structural components and recognition aims to recognize individual characters/words.

 

Related papers:

      "Pink Panther: A Complete Environment for Ground-truthing and Benchmarking Document Page Segmentation", Berrin Yanikoglu and Luc Vincent, Pattern Recognition 31, 9 1998.

 

Papers in Turkish:

      “Belge Bölütleme ve Tanıma”, Türkiye Bilişim Ansiklopedisi, 2006

 

Thesis supervised:

  • Esma F. Bilgin, PhD student
  • Damla Arifoğlu, PhD student

 

 

 

 

 

 

Protein Folding

 

Finding the unknown 3D shape of a protein, from its known aminoacid sequence (1D) is a major research problem. We have concentrated on the subproblems of finding the 2D structure and estimating contact maps.

 

Related papers:

·        “Protein Structural Class Determination Using Support Vector Machines”, Zerrin Isik, Berrin Yanikoglu, Ugur Sezerman, to appear in Lecture Notes in Computer Science-ISCIS 2004, Oct. 2004.

·       "Minimum Energy Configurations of the 2-Dimensional HP-Model of Proteins by Self Organizing Networks", Berrin Yanikoglu and Burak Erman, Journal of Computational Biology, vol:9, no:4, 2002.

·       "Fold Classification Using Support Vector Machines", Zerrin Işık, Uğur Sezerman, Berrin Yanikoglu, Protein Structure Prediction Workshop, November 1-3 2002, Antalya,Turkey

 

 

 

 

PhD Students

(Past and Current)

 

 

 

 

 

Alisher Kholmatov, Privacy Preserving Biometric Authentication Systems, PhD thesis, Jan. 2008, Sabanci University.

 

Alisher’s work Berkay’s work is the state-of-art online (tablet based) signature verification system in the world, having won most of the international competitions sin the area. 


Most cited paper award, Pattern Recognition journal, 2005-2010.

Berkay Yılmaz, Offline Signature Verification with User-Based and Global Classifiers of Local Features, PhD thesis, March. 2015, Sabanci University.

 

Berkay’s work is one of the state-of-art offline (image) signature verification systems in the world, having won several of the international competitions şn the area. 


Çağlar Tırkaz, Sketch and Attribute Based Query Interfaces, PhD thesis, May 2015, Sabanci University.

Co-supervised by T. Metin Sezgin, Koç U.

Muhammet Yıldız

 

I graduated from Bilkent University Computer Engineering department in 2006. The same year I joined TUBITAK UEKAE as a Researcher. In 2010, I got my MS degree from GYTE degree with my thesis titled "Combining Turkish e-Id card with online credit card Transactions".

 

My PhD topic is on biometric privacy and security, and face and fingerprint recognition.

 

 

Rahim Dehkharghani

 

I am a computer engineer who is  M.Sc graduated of Computer Engineering from Shahid  Beheshti university of Iran and now is studying Computer Science and Engineering at Sabanci university, Istanbul, Turkey. My research interest is Knowledge Engineering and Natural Language Processing. Also I am interested in Machine learning and Data Mining. My M.Sc. thesis subject was Bilingual Ontology Mapping which caused to publish a book named bilingual ontology mapping which was the first one in this area. I was a member of the Natural Language Processing Laboratory of Shahid Beheshti University and active in Farsnet project (building a Wordnet for Persian language).

 

My PhD thesis topic is information retrieval and opinion mining using Machine learning and NLP techniques.

 

Co-Supervised by Yücel Saygın.

 

 

Esma Bilgin

 

Thesis topic: Online handwriting recognition using HMM and Depp learning.

 

Note: Maybe a coincidence or not J, my later PhD students are all female. J

Damla Arifoğlu

 

Thesis topic: Online handwriting recognition & retrieval.

 

Gizem Gezici

 

Thesis topic: Sentiment Analysis for Turkish.

Ece Egemen

 

Thesis topic: Information Retrieval for Turkish.

Current MS Students

 

 

Tolga Yıldıran

Mostafa Mehdipour Ghazi