Frequency based approach in multimodal biometric fusion for personal authentication

Abstract

In this research, a novel multimodal biometric system was developed that combined various unimodal biometric traits like fingerprint, palmprint, knuckleprint and retina image inputs. The multimodal system used feature-level fusion to generate the fused template by combining various biometric traits. Gabor filter was employed to extract the features from the different biometric traits. The extracted features from the fingerprint, knuckleprint, palmprint and the retina were fused together as one template and stored in the database for authentication, thus reducing spatial and newlinetemporal complexity. Several experiments were carried out with sample images obtained from different public databases like FVC2000, FVC2002 and FVC2004 fingerprint databases, Hong Kong Polytechnic University (PolyU) knuckleprint image database, CASIA palmprint image database and the Fundus and the Drive retina image databases. As a contribution of the research, a novel system for protecting newlinefingerprint privacy by combining three separate fingerprints of thumb, index and middle fingers and enrolled as a new template. An innovative approach for Fingerprint Feature extraction by using the Fixed-Size Template (FEFST) method was proposed. Minutiae positions (ridge endings, ridge bifurcation) and orientations with the reference points were extracted from all the fingerprints. The fingerprint which has the maximum ridge endings was taken as the main template. To this template the templates of the other two fingerprints was embedded to generate the combined minutiae template. The generated combined minutiae template was stored in a database for enrollment. In the authentication stage, the three query fingerprints were fed into the system and matched against the stored template. newline newline