Feature level fusion of face and fingerprint modalities using Gabor filter bank

This paper proposes a biometric authentication system based on feature level fusion of face and fingerprint modalities. The proposed method utilizes Gabor filter bank with two scales and eight orientations, to extract directional features from source data. Usage of a small set of Gabor filters typically reduces the system processing time. To introduce a good discriminating ability and to avoid curse of dimensionality in feature set, we have used Principal Component Analysis (PCA) +Linear Discriminant Analysis (LDA) framework. The framework enables us to use only 39 features as input to classifier stage of the system. Distance classifiers are employed to authenticate a person based on the distance between input image and stored database templates. Experimental results showcase the advantages of feature level fusion over a uni-modal framework. The system achieves recognition accuracy up to 99.25%. The experiments have been carried out on ORL face database and FVC2002 fingerprint database.