Título: Reconhecimento Off-line de Assinaturas Utilizado MLP-Backpropagation e Momentum: um Estudo Comparativo
Autores: Gomes, Herman Martins; Carvalho Filho, Edson Costa de Barros
Resumo: One of the most problems with off-line signature recognition is the drastical reduction of useful information due to the fact that all dynamic features are reduced to a single statical image. A great variety of promissing techniques. That have been established as apropriated to general tasks of patern recognition, may be used in off-line signature recognition to better the performance, like Neural Networks and Momentum. The main goal of this paper is to discuss and to compare Neural Networks (represented by MLP-Backpropagation paradigm) and Moment-based techniques (represented by Standard Moments) in off-line signature recognition. We start with a brief introduction to the off-line signature recognition problem. Afterwards, we give a description of the investigated techniques. At the end, a signature database, the design of experiments and results are presented, and conclusions related.
Código DOI: 10.21528/CBRN1994-021
Artigo em PDF: CBRN1994-paper21.pdf
Arquivo BibTex: CBRN1994-paper21.bib