Research • Research Areas

The information-driven societies we live in are becoming increasingly dominated by information-bearing signals. Such signals include speech, audio, camera images, video, biomedical signals and images, satellite images, gene expression data, transmitted signals in wired and wireless networks, as well as radar images, and so on. While we are reasonably capable of measuring and recording such data by advanced sensors, the state of our technology is still in its infancy in terms of intelligent processing and interpretation of the signals. Yet, considerable demand exists for such automated information extraction tools in many tasks that currently require laborious manual work or an inefficient human-machine interface. For example,
imagine the possibility of medical imaging systems that not only provide an image, but also mark all anatomical structures, detect possible anomalies, and suggest possible diagnoses to aid the physician. Alternatively, picture a computer that you can communicate with using speech and gestures rather than a keyboard and mouse.

At Sabancı University, a group of researchers develop new methods for the effective and efficient extraction of information from signals for a variety of applications. The following three main areas of research are representative of the current projects:

Signal processing:

Source localization using acoustic sensors; methods for the automated analysis of electroencephalogram (EEG) signals from the brain; statistical inference algorithms for wireless sensor networks.

Image and video processing:

Advanced radar imaging techniques from data collected by unmanned air vehicles; ultrasound imaging techniques for the nondestructive evaluation of materials; algorithms for the segmentation of anatomical structures in medical images; video-based statistical inference techniques for driver fatigue detection.

Speech and pattern recognition:

Acoustic and language modeling techniques for large vocabulary continuous speech recognition; discriminative feature extraction and dimensionality reduction techniques for robust pattern recognition; machine learning methods for biological sequence analysis particularly for secondary structure prediction.

Medical Image Analysis and Computer Vision:

Computer Aided Diagnosis, Monitoring Methods, Content Extraction from Medical Images, 3D Shape Analysis and Modeling,  3D Reconstruction from 2D images.

Faculty: Mujdat Cetin, Hakan Erdogan, Aytul Ercil, & Gozde Unal

Recent Projects:

-- Signal, Image Processing and Pattern Recognition for Intelligent Automation Center, supported by the European Commission through a Specific Support Action Grant (2005- 2008).

-- DRIVE-SAFE: Signal Processing and Advanced Information Technologies for improving Driver/Driving Prudence and Accident Reduction, supported by the State Planning Organization of Turkey (2005-2007).

-- MR-based Analysis, Indexing, and Retrieval of Brain Iron Deposition in Basal Ganglia, supported by the European Commission through a Transfer of Knowledge Grant. (In collaboration with Philips Research, the Netherlands) (2006- 2008).

-- New Generation Information Processing Techniques for Imaging Sensors and Wireless Sensor Networks, supported by the European Commission through a Marie Curie Reintegration Grant (2006-2008), and by TUBITAK through a Career Award (2006-2011).

-- DBSP-NEDO: Driving Behavior Signal Processing, supported by the New Energy and Industrial Technology Development Organization (NEDO), Japan (2005-2008).

 

Recent/Relevant Papers:

-- Region-enhanced passive radar imaging, Cetin, M., Lanterman, A., IEE Proc. Radar, Sonar & Navigation, 152(3), 185-194, 2005.

-- A sparse signal reconstruction perspective for source localization with sensor arrays, Malioutov, D. M., Cetin, M., Willsky, A. S., IEEE Trans. Signal Processing, 53(8), 3010-3022, 2005.

-- A nonparametric statistical method for image segmentation using information theory and curve evolution, Kim, J., Fisher, J., Yezzi, A., Cetin, M., Willsky, A.S., IEEE Trans. Image Processing, 14(10), 1486-1502, 2005.

-- Multimodal person identification for human vehicle interaction, Erzin, E., Yemez, Y., Tekalp, A. M., Ercil, A., Erdogan, H., Abut, H., IEEE Multimedia, 13(2), 18-31, 2006.

-- Radar image of a backhoe loader formed by an advanced algorithm, developed by Sabancı University researchers.

-- 2-D robust recursive least squares lattice algorithm and its application to defect detection in textured images, Meylani, R., Ertuzun, A., Ercil, A., IEICE Transactions, 2006.

-- Using semantic analysis to improve speech recognition performance, Erdogan, H., Sar>kaya, R., Chen, S. F., Gao, Y., Picheny, M., Computer Speech and Language, 19(3), 321-343, 2005.

-- Semantic confidence measurement for spoken dialog systems, Sarikaya, R., Gao, Y., Picheny, M., Erdo¤an, H., IEEE Trans. Speech and Audio Processing, 13(4), 534-545, 2005.

-- [2008] - Unal, Gozde, Slabaugh, Greg, Estimation of Vector Fields in Unconstrained and Inequality Constrained Variational Problems for Segmentation and Registration, Journal of Mathematical Imaging and Vision.

-- [2008] - Gozde Unal, S. Bucher, S. Carlier, G. Slabaugh, Shape-driven Segmentation of the Arterial Wall in Intravascular Ultrasound Images, IEEE Trans. On Information Technology in Biomedicine.

-- [2007] - Gozde Unal, Anthony Yezzi, Stephano Soatto, and Greg Slabaugh, A Variational Approach to Problems in Calibration of Multiple Cameras, IEEE Transactions on Pattern Analysis and Machine Intelligence (PAMI), Vol. 29, No. 8, August, 2007, pp. 1322-1338/.