Maximum Entropy Estimation of the Probability Density Function from the Histogram Using Order Statistic Constraints: Applications in Cow ovulation and Earth Surface Classification

Past

Abstract: Many practical problems involve fundamental detection theory. Often the fundamental probability densities of the features of the events are unknown. Two examples we have worked with in recent years are 1) dairy cow body temperature events (ovulation, birthing, etc.). The cow state is a transient signal whose features’ densities must be learned and optimally detected. Once the statistics are “known”, classical detectors or classifiers can be applied. 2) earth surface classification from radar reflections. The earth surface problem is treated as approximately 2-dimensional stationary, but the densities must first be estimated before classifiers can be constructed. Speaker: Dr. Rodney L. Kirlin Dr. Kirlin obtained his bachelor’s and master’s degrees from University of Wyoming, and a doctoral degree from Utah State University. He is a researcher and consultant with problem-solving and algorithm development skills applicable to statistical detection, estimation and presentation within diverse applications typical of communications sonar, seismology and radar. He has consulted in several industries for both private and government organizations, including Amoco and the US and Canadian Navies. Dr. Kirlin has published over 70 refereed journal papers and 200 consulting contract reports. He is a Fellow of the IEEE. For further information, please contact Prof. Kin Fun Li, Electrical and Computer Engineering, University of Victoria, (kinli@uvic.ca)