Bimodal Biometrics for Efficient Human Recognition Using Wavelet, Principal Component Analysis and Artificial Neural Network

This paper proposes an effective fusion scheme for extracting more discriminative information from bimodal biometrics at data, feature and decision levels. In all these three levels of fusion, information from both face and fingerprint image of a single subject are fused to effectively represent it in a more discriminative ways. For all these three approaches, a combination of wavelet and principal component analysis for feature extraction steps and multi-layer perceptron MLP neural network for classification is applied. Experimental results show the feasibility of the proposed method. Results show that Face and Finger print based bimodal biometrics achieve better EER, FAR and FRR than that of the unimodal biometrics.