HIDDEN MARKOV MODEL BASED WEIGHTED LIKELIHOOD DISCRIMINANT FOR MINIMUM ERROR SHAPE CLASSIFICATION (FriPmOR6)

Author(s) : Ninad Thakoor (University of Texas at Arlington, United States of America) Jean Gao (University of Texas at Arlington, United States of America) Sungyong Jung (University of Texas at Arlington, United States of America)

Abstract : The goal of this communication is to present a weighted likelihood discriminant for minimum error shape classification. Different from traditional Maximum Likelihood (ML) methods in which classification is made based on probablities from independent individual class models as which is the case for general hidden Markov model (HMM) methods, our proposed method utilizes information from all classes to minimize classification error. Our proposed approach uses a Hidden Markov Model as a curvature feature based 2D shape descriptor. To achieve a discrimant function with minimum classification error, in this contribution we present a Generalized Probabilistic Descent (GPD) method to weight different curvatures. In contrast with other approaches, a thus weighted discrimant function is introduced. We believe our sound theory based implementation reduces classification error by combining hidden Markov model with generalized probablistic descent theory. We show comparison results obtained with our approach and classic maximum-likelihood calculation for fighter planes in terms of both discrimant functions and classification accuracies.

Menu