Rank Level Integration of Face Based Biometrics

This paper investigates the integration of two modalities: facial thermograms and ear, extracted from the same face, by using rank level fusion scheme. The first modality consists of the infrared thermal faces acquired using infrared camera whereas the second one constitutes point features on the ear imaged using ordinary digital camera. The acquired facial thermo grams and ear images are first normalized by locating ROI and then features are extracted using Haar wavelets and SHIFT (Scale Invariant Feature Transform) respectively. Integration of their associated ranks has been done by using the modified Borda count and logistic regression methods. The proposed authentication system is tested on 500 facial thermo grams and ear images and operates on 98% of genuine acceptance rates (GAR) at 0.1% of false acceptance rate (FAR). Although substantial work remains to be done, yet our results indicate that the rank level integration of facial thermo grams and ear images is poised to provide a promising direction to the face based multimodal biometric systems.