SCORE FUSION OF FINGER VEIN AND FACE FOR HUMAN RECOGNITION BASED ON CONVOLUTIONAL NEURAL NETWORK MODEL

Abstract

In the last decade, the biometrics refers to automatic recognition of persons using their physiological or behavioral characteristics. The combination of multiple biometrics or, multimodal biometrics have higher accuracy to verify the person and ensure that its information or data is safer compared to system based on single biometrics modality. In this regard, this paper introduces a scheme for multimodal biometric recognition system based on the fusion of finger-vein and face images using Convolutional Neural Network (CNN) and different classifiers. The pre-processed finger-vein image using Adaptive Histogram Equalization (AHE) is input into a CNN model. Then, Random Forest (RF) classifier performs as a recognizer. In addition, a hybrid CNN-Linear Support Vector Machine (SVM) model is used for recognizing face images. After this process, the score level fusion of bimodal biometric based on the weighted concatenation is applied to identify the identity of the individual. Experimental results on publicly available VERA Fingervein database, Color Feret and Ar face database have shown significant capability of identification biometric system. The proposed system provides high recognition accuracy rate by 99,98% compared with other classical methods and traditional techniques based on normal recognition or CNN architectures.