Home  | Legals | Sitemap | KIT

A Robust Face Recognition Algorithm for Real-World Applications

We developed a local appearance-based face recognition algorithm using discrete cosine transform, which is a generic, robust, and fast face recognition algorithm that has been deployed for several real-world person identification applications. The proposed face recognition approach divides the input face image into local blocks and processes each local block using discrete cosine transform. The local representation provides robustness against appearance variations in local regions caused by factors such as facial occlusion or expression, whereas utilizing frequency information provides robustness against changes in illumination. The algorithm has been extensively tested both using standard benchmark databases —AR, CMU PIE, FRGC, Yale B, Extended Yale B— and using the data collected from real-world applications —person identification in smart rooms, entrance monitoring, visitor interface, person re-identification in TV series—. The experimental results show that, the algorithm can successfully handle facial appearance variations caused by uncontrolled recording conditions, expression, occlusion, and illumination. Moreover, the systems based on this algorithm have been found to work reliably under real-world conditions.
Related Publications
Author Title Source

H.K. Ekenel, R. Stiefelhagen

In Proc. of CVPR Biometrics Workshop, New York, USA, June 2006.

H.K. Ekenel, R. Stiefelhagen

In Proc. of Intl. Conf. on Pattern Recognition (ICPR’10), Istanbul, Turkey, August 2010.

H.K. Ekenel, R. Stiefelhagen

In Proc. of CVPR Biometrics Workshop, New York, USA, June 2006.

H.K. Ekenel, R. Stiefelhagen

In Proc. of Intl. Conf. on Biometrics: Theory, Applications and Systems (BTAS’09), Washington DC, USA, September 2009.

H.K. Ekenel, R. Stiefelhagen

Intl. Conf. on Biometrics (ICB’09), LNCS, Vol. 5558, pp. 367-375, Alghero, Italy, June 2009.

H.K.Ekenel, R. Stiefelhagen

In Proc. of 13th European Signal Processing Conference (EUSIPCO 2005), Antalya, Turkey, September 2005.

H. Gao, H.K. Ekenel, M. Fischer, R. Stiefelhagen

In Proc. of Intl. Conf. on Pattern Recognition (ICPR’10), Istanbul, Turkey, August 2010.

H. Gao, H.K. Ekenel, R. Stiefelhagen

Intl. Conf. on Biometrics (ICB’09), LNCS, Vol. 5558, pp. 32-41, Alghero, Italy, June 2009.

H.K. Ekenel, R. Stiefelhagen

Intl. Conf. on Biometrics (ICB’09), LNCS, Vol. 5558, pp. 299-308, Alghero, Italy, June 2009.



Related Theses
Author Title Date
Hazım Kemal Ekenel 2009/02


Identifying Subjects in a Smart Room

This system is deployed at a seminar room. Four cameras are mounted at the corners of the room. Identification is done using video-streams acquired from these four cameras.

Related Publications
Author Title Source

K. Bernardin, H.K. Ekenel, R. Stiefelhagen

Personal and Ubiquitous Computing 13(1): 25-31, 2009.

A. Pnevmatikakis, H.K. Ekenel, C. Barras, J. Hernando

Book Chapter in Computers in the Human Interaction Loop, Human-Computer Interaction Series, Springer Verlag, London, 2009.

R. Stiefelhagen, K. Bernardin, H.K. Ekenel, M. Voit

In Proc. of 8th IEEE Int. Conference on Face and Gesture Recognition, Amsterdam, Netherlands, September 2008.

H.K. Ekenel, J. Stallkamp, H. Gao, M. Fischer, R. Stiefelhagen

In Proc. of IEEE International Conference on Multimedia & Expo, pp. 1007-1010, Beijing, China, July 2007.

H.K. Ekenel, M. Fischer, R. Stiefelhagen

Machine Learning for Multimodal Interaction, LNCS, Vol. 4892, pp. 120-131, Brno, Czech Republic, June 2007.

H.K. Ekenel, M. Fischer, Q. Jin, R. Stiefelhagen

In Proc. of CVPR Biometrics Workshop, Minneapolis, USA, June 2007.

H.K. Ekenel, Q. Jin, M. Fischer, R. Stiefelhagen

CLEAR Evaluation Workshop, LNCS, Vol. 4625, pp. 256-265, Baltimore, US, May 2007.

R. Stiefelhagen, K. Bernardin, H.K. Ekenel, J. McDonough, K. Nickel, M. Voit, M. Wölfel

Signal Processing, Vol. 86 (12), Dec. 2006.

H.K. Ekenel, A. Pnevmatikakis

In Proc of 7th International Conference Automatic Face and Gesture Recognition (FG2006), Southampton, UK, April 2006.

H.K. Ekenel, Q. Jin

CLEAR Evaluation Workshop, LNCS, Vol. 4122, pp. 249-257, Southampton, UK, April 2006.



Recognizing Individuals Entering a Room

The system is deployed at the entrance door of a seminar room. The camera is located opposite the door at a distance of six meters. Individuals are recognized automatically when they enter the room.


Related Publications
Author Title Source

H.K. Ekenel, J. Stallkamp, R. Stiefelhagen

Computer Vision and Image Understanding, Vol. 114, No. 5, pp. 596 – 608, May 2010.

J. Stallkamp, H.K. Ekenel, R. Stiefelhagen

In Proc. of International Conference on Computer Vision (ICCV'07), pp. 1-8, Rio de Jenario, Brasil, October 2007.

H.K. Ekenel, J. Stallkamp, H. Gao, M. Fischer, R. Stiefelhagen

In Proc. of IEEE International Conference on Multimedia & Expo, pp. 1007-1010, Beijing, China, July 2007.

H.K. Ekenel, R. Stiefelhagen

In Proc. of International Conference on Computer Vision Systems, ICVS 2007, Bielefeld, Germany, March 2007.



Face Recognition for Humanoid Robots

The system is deployed on a robot and uses a stereo camera for image acquisition. It identifies the person interacting with the robot.


Related Publications
Author Title Source

H. Holzapfel, T. Schaaf, H.K. Ekenel, C. Shaa, A. Waibel

In Proc. of 29th German Conference on Artificial Intelligence (KI2006), Bremen, Germany, June 2006.

K. Nickel, H.K. Ekenel, M. Voit, R. Stiefelhagen

In Proc. of 2nd Intl. Workshop on Human-Centered Robotics Systems, Munich, October 2006.

S. Könn, H. Holzapfel, H.K. Ekenel, A. Waibel

In Proc. of International Conference on Computer Vision Systems, ICVS 2007, Bielefeld, Germany, March 2007.

R. Stiefelhagen, H.K. Ekenel, C. Fügen, P. Gieselmann, H. Holzapfel, F. Kraft, K. Nickel, M. Voit, A. Waibel

IEEE Transactions on Robotics, Special Issue on Human-Robot Interaction, Vol. 23, No. 5, October 2007.



3D Face Recognition

In the system, 3D point clouds are registered to provide dense correspondence between faces. Depth images are constructed from the corresponding well-registered point clouds. The system utilizes depth map images to extract local features and performs identification using local appearance-based face recognition.

Related Publications
Author Title Source

H. K. Ekenel, H. Gao, R. Stiefelhagen

IEEE Transactions on Information Forensics and Security, Vol. 2, No. 3, pp. 630-635, September 2007.



Open-set Face Recognition

In the system, faces are automatically detected and registered. The local appearance-based face recognition method is utilized for representing the faces. An identity verification component is trained for every known subject in the database. Open-set identification is performed via a series of verification processes.

The system has been developed as a visitor interface, where a visitor looks at the monitor before knocking on the door. A welcome message is displayed on the screen. While the visitor is looking at the welcome message, the system identifies the visitor unobtrusively without needing person’s cooperation. According to the identity of the person, the system customizes the information that it conveys about the host.


Related Publications
Author Title Source

H.K. Ekenel, L. Szasz-Toth, R. Stiefelhagen

7thIntl. Conf. on Computer Vision Systems, LNCS, Vol. 5815, pp. 43-52, Liege, Belgium, October 2009.



Identity-based Interactive Video Retrieval

The system automatically detects and tracks persons in video sequences of TV series, and extracts features from these tracks which can be used to reliably identify the persons in the video.

In the application, the user selects a person from a scene and the system returns scenes in the video that contain the same person. The user can refine the search interactively.


Related Publications
Author Title Source

M. Fischer, H.K. Ekenel, R. Stiefelhagen

In Proc. of Intl. Workshop on Content-based Multimedia Indexing (CBMI’10), Grenoble, France, June 2010.



Driver Identification in Smart Cars

Car Driver
(Source: flickr.com)
The system identifies the driver of the car. Face analysis methods can also be used to provide driver assistance.

 

Related Publications
Author Title Source

J. Stallkamp, H.K. Ekenel, H. Erdogan, R. Stiefelhagen, A. Ercil

Workshop on DSP in Mobile and Vehicular Systems, Istanbul, Turkey, June 2007.

H. Erdogan, A. Ercil, H.K. Ekenel, S.Y. Bilgin, I. Eden, M. Kirisci

6th International Workshop on Multiple Classifier Systems (MCS 2005), LNCS, Vol. 3541, pp. 366–375, California, USA, June 2005.