Title :
Recognition of character using morphological associative memory
Author :
Akyama, Marcio Teruo ; Kikuti, Marcio
Author_Institution :
Instituto de Computacao, Univ. Estadual de Campinas, Sao Paulo, Brazil
Abstract :
The article provides a research in pattern recognition using a hardware implemented classifier based on morphological associative memory (MAM). Most neural network (NN) models used for this purpose have a complex learning rule. Instead, we propose a morphological classifier that has a simple learning rule based on mathematical morphology and presents an infinity storage capacity. Initially, we define a standard for the image as pre-requirement for pre-processing. The pre-processing stage is performed to reduce the dimensionality of MAM´s data input in order to act as an interface between the inputs and the classifier. A practical application on recognizing characters from the Roman alphabet shows the versatility of this architecture
Keywords :
content-addressable storage; mathematical morphology; optical character recognition; pattern classification; MAM; NN models; Roman alphabet; character recognition; complex learning rule; data input; hardware implemented classifier; infinity storage capacity; mathematical morphology; morphological associative memory; morphological classifier; neural networks; pattern recognition; pre-processing stage; simple learning rule; Algebra; Associative memory; Character recognition; Field programmable gate arrays; H infinity control; Hardware; Morphology; Neural networks; Optical character recognition software; Pattern recognition;
Conference_Titel :
Computer Graphics and Image Processing, 2001 Proceedings of XIV Brazilian Symposium on
Conference_Location :
Florianopolis
Print_ISBN :
0-7695-1330-1
DOI :
10.1109/SIBGRAPI.2001.963106