Seminar by Krithika Venkataramani
Designing Classifier Ensembles for Accurate Classification
Krithika Venkataramani
Tata Consultancy Services, Hyderabad
Date: Monday, September 7, 2009
Time: 11:00 AM
Venue: CS101.Abstract:
Biometric authentication is being considered for secure access to physical and virtual spaces in place of techniques employing cards and passwords since biometrics cannot be lost or stolen. Typical practical applications have a large range of natural variability in biometric features. Pose, illumination and occlusion present major challenges in face recognition. Challenges in fingerprint recognition are rotation of the finger with respect to the sensor, elastic distortion caused by pressing the finger on a sensor surface, and varying environmental conditions such as dryness, moisture, dirt, etc. A set of classifiers help in handling the different variations well. The theory of generating the set of classifiers in an optimal manner, in terms of accuracy for a given number of classifiers is presented. Different ways of generating the classifier ensemble for various databases are presented. This theory is applicable to many other pattern recognition problems where a large variability in data is present. Applications to digital watermarking in images and video are demonstrated.
About The Speaker:
Dr. Krithika Venkataramani obtained her B. E. degree in Electronics and Communication from I. I. T. Roorkee in 2000, and her M.S. and Ph.D. degrees from the Electrical and Computer Engineering Department at Carnegie Mellon in 2002 and 2007, respectively. She is currently working with Tata Consultancy Services at Hyderabad. She has authored or co-authored 2 book chapters and over 15 journal and conference papers. Her current research interests include pattern recognition, computer vision, signal and image processing, biometric recognition, video surveillance and digital watermarking. Back to Seminars in 2009-10