Feature Level Fusion in Multimodal Biometric Authentication System

Abstract

Multimodal systems integrate multiple sources of human information to ensure high level security. Multimodal biometric systems increase the recognition rate of the biometric systems either by reducing the false acceptance rate (FAR) or false rejection rate (FRR). Multiple biometric traits can be combined at feature level. Feature level fusion increases the reliability of the system by preventing the biometric template from modification. In the proposed system, feature level fusion is employed to fuse the feature vectors of iris and ear extracted by Principal Component Analysis technique, which also reduces the dimension of the feature vectors. Finally matching is performed by comparing the test fused feature vectors with all training images using distance measure. This system is developed to study and analyze, whether the performance of multimodal biometric system is improved over unimodal biometric system by attaining 93% success rate when fusion is inclined.