ComposiÃÃo de biometrias para sistemas multimodais de verificaÃÃo de identidade pessol

AUTOR(ES)
DATA DE PUBLICAÇÃO

2005

RESUMO

This work presents contributions for the problem of identity verification thought an architecture that combines face, signature and keystroke dynamics. The first two biometrics were chosen based on the fact that they are already integrated in the society, besides the equipment to capture these modalities are cheap. The third biometrics chosen, keystroke dynamics, is also cheap and transparent to the user. The main motivation of this thesis is to analyze combination strategies aiming to improve the performance of personal identification systems. Therefore, it was identified and investigated the following points: (i) Signature verification of different sizes using seven features groups: pseudo-dynamics, structural, and invariant (moments of Hu, Maitra, Flusser, Simon and Central); (ii) Face classification through Eigenbands Fusion (a face image was divided into bands and the best principal components of each band was assembled to form the feature vector); (iii) Keystroke dynamics verification through the analyzes of press and latency time of the users; (iv) Architecture arrangement aiming to combine three biometrics, besides experiments to measure performance; (v) Local threshold investigation to separate regions of authentic and impostor users through t-Student distribution. The results showed that the combined system over performed each one of the biometrics modalities when used separately

ASSUNTO(S)

signature frication local threshold cÃlculo de limiar local verificaÃÃo de assinaturas multimodal personal identification identificaÃÃo pessoal multimodal combinaÃÃo de decisÃo face classicafition ciencia da computacao keystroke dynamics verification decision combination classificaÃÃo de faces verificaÃÃo atravÃs de dinÃmica da digitaÃÃo

Documentos Relacionados