IEEE Seminar Announcement

Abstract: Often automatic target recognition systems must be designed with relatively small amount of training data. In addition, practical design requires small computational complexity. Plug-in test statistics suffer from large estimation errors, often causing performance to degrade with increasing size of the measurement vector. Choosing a better test statistic or applying a method of dimensionality reduction are two possible solutions to the problem above. In this presentation, we consider a recognition problem where the data for each population are modeled as being complex Gaussian with zero mean and unknown variances. The system is designed to implement a plug-in log-likelihood ratio test with maximum likelihood estimates of the unknown parameters in place of the true parameters. Because a small amount of data is available to estimate the parameters, the performance of such a system is strongly degraded relative to the performance with known parameters. To improve the performance of the system we define a thresholding function that, when incorporated into the plug-in log-likelihood ratio, significantly decreases the probability of error for binary and multiple hypothesis testing problems. We analyze the modified test statistic and present the results of Monte-Carlo simulations.

Speaker Bio: Natalia Schmid received the M.S. in applied mathematics and physics from Moscow Institute of Physics and Technology in 1991. She received the candidate degree from Russian Academy of Science in 1995 and her D.Sc. in electrical engineering from Washington University in St. Louis in 2000.
From 2001 to 2002 she held a postdoctoral position at the University of Illinois at Urbana-Champaign. In 2003 she joined the faculty in the Department of Computer Science and Electrical Engineering at West Virginia University. Her research interests include stochastic image processing, stochastic recognition, and authentication with applications to biometrics, medical imaging, radar imaging, and computer vision problems. Her primary focus is on mathematical modeling and performance analysis. Current research projects include: large deviations performance analysis for biometrics recognition; CT imaging in the presence of objects with unknown shape; and regularization methods for recognition problems.