| The use of biometrics has been proposed over traditional identity verification systems, such as login and password, since biometrics are more difficult to steal, forge, or transfer. Fingerprints is a physiological biometric that is unique, stable over time, non-intrusive and can be reliably measured using a fingerprint scanner. The goal of this thesis is to design and implement a fingerprint verification algorithm that complies with strict execution speed, memory space and accuracy requirements measured in False Rejection Rate (FRR) and False Acceptance Rate (FAR). Although accuracy and memory requirements were evaluated using a prototype developed in MATLAB™, the prototype considered the architectural features of the target implementation platform, the coarse-grained parallel processor TMS320C80, so that it could comply with the speed requirements. Although the algorithm did not reached the specified accuracy requirements (FRR < 5% and FAR < 0.1%), the results (FRR = 20% and FAR = 5%) validate the implemented approach as a baseline for future improvements. |
