Autenticação Pessoal por Imagens de Sinais Gráficos

Abstract

In this work is proposed a personal authentication method based on images composed of handwritten strokes that may represent symbols, words, signatures or any kind of drawings. This method is efficient, robust and can be easily implemented, requiring low CPU processing power and needs only five handwritten signal samples for enrollment. Those issues allow the proposed method to be suitable for real time authentication systems and control access applications. The authentication method has been evaluated under both verification and identification processes. In verification experiments random and skilled forgeries were used, while in identification experiments three different database configurations were setup. Performance of 0.7 % equal error rate and that of 93.7 % correct classification rate were achieved in the verification and identification processes, respectively.