Multimodal biometric fusion based on fingerprints and faces

Žiga Emeršič (2013) Multimodal biometric fusion based on fingerprints and faces. EngD thesis.

Preview

PDF Download (4Mb)

Abstract

The diploma thesis presents the procedure of multimodal biometric fusion, which combines the verification system based on fingerprints and the verification system based on faces. The fusion is implemented at the matching score level, namely the fusion based on fusion rules and the fusion based on classification methods. For the fingerprint's part of the input, the FingerIdent's output is used. On the basis of which, the matrix of distances is then created and used in the fusion. Both systems work on the learn and the test set. Each consists of 800 samples - 8 samples per person, 100 persons total. The size of the input matrices used for the fusion is 800x800. The performance of the fusion was addressed through the following measures: ROC, AUC, EER, Verification at 0,1% FAR. The results have shown that the fusion improves the performance of the system.

Actions (login required)