Título: Reconhecimento de Padrões através de uma Rede Neural Artifical Baseada em Mecanismos da Visão
Autores: Rodrigues, Marco Aurélio Benedetti; Rodrigues, Iria Pedroso da Cunha; Marino Neto, José; Azevedo, Fernando Mendes
Resumo: While a variety of algorithms and techniques for pattern recognition has been developed, the vertebrate eye and visual system remains undisputedly considered as one of the better biological mechanism capable of accurate pattern discrimination and identification. In the present study, we advance an artificial neural network model, based on some relevant structural and functional characteristics of the human retina. This model was evaluated by assessing its capabilities in detecting K-complexes of human eletroencephalographic recordings, comparing its results with those of an medical specialist on this field to determine sensibility and specificity attributes of the system. Our results show that such an approach brings about a number of favorable attributes, as compared to other and more usual ones, including minimization of false-positive detection due to the double identification process by “foveal” and “extra-foveal” circuits.
Código DOI: 10.21528/CBRN2003-088
Artigo em PDF: 6CBRN_088.PDF
Arquivo BibTex: 6CBRN_088.bib