No abstract

F~EtesamiThis thesis describes an approach for accomplishing a pattern recognition task using conceptual graph theory and neural networks (NNs). The set of patterns to be recognized are the capital letters of six different fonts of the English alphabet, plus two shifted and six rotated versions of each. The letters are represented to the neural network on a 16x16 input grid (256 "sensor lines"). A standard classification encoding for such patterns is to use a 26-bit vector (26 lines at the NN' s output), one bit corresponding to each letter. Experiments with such an encoding yielded results with poor generalization capability. A new encoding scheme was developed, based on the conceptual graph formalism. This entailed designing a set of concepts and a set of associated relations 2 app:.'"Opriate to the upper case letters of the English alphabet. From these, the following were developed: a conceptual graph representation for each letter, a connection matrix for each, and finally, a C-vector and an R-vector representation for each. The latter were used as the output encoding (21 bits) of the NN pattern recognizer. A feed-forward neural network with 256 inputs, 21 outputs, and 2 hidden layers was trained using the backpropagation-of-error algorithm. Results were significantly better than with the more standard. encoding. Generalization on fonts improved from 74% to 96%, generalization on rotations improved from 83% to 94%, and finally, generalization on shifts improved from 2% to93%. ACKNOWLEDGEMENTSThis research project could not have been made possible without the tremendous amount of help of Dr. George G. Lendaris. In particular, his careful assistance in the development of thoughts and experiments has made the production of this thesis run much more smoothly than I have ever hoped; I owe him my deepest respect and a particular debt of gratitude. I appreciate as well the contribution of Dr. Faris Badi'i on data preprocessing methods for pattern recognition applications. Again, I wish to thank my friends Mohammad Assaf and Shihab Hanayneh for the enthusiasm they showed, and the comments they gave.Finally, I would like to thank my family for their suppon, understanding, and encouragement throughout my years of study, helping give in the process a very special meaning to the creation of this thesis. PATTERN RECOGNITION DefinitionPattern recognition (by machine) is typically considered as the categorization of input patterns into their respective classes via feature extraction, wherein an individual pattern is characterized by the relations among its constituent features, rather than by the original measurements via which it was acquired [Wiener, 1986]. The following is a 2 typical definition of the pattern recognition problem: there exists a set of N objects divided into M nonintersecting subsets, referred to as object classes (for the problem of this thesis, alphabetic characters); to each object, there corresponds a particular description U; it is required to implement a system which, on the basis of such a des...