Fusion of Finger Vein Images, at Score Level, for Personal Authentication

Abstract

A biometric system with a single biometric trait is less effective, owing to constraints, such as inter-class similarities, susceptibility to noisy pictures and spoofing. Integrating information from different biometric evidences aids in the resolution of difficulties in unimodal biometric systems. It is incredibly challenging in a biometric system to intrude into more than one trait at the same time. Researchers are becoming more interested in multimodal biometric systems due to benefits such as dependability, security, and robustness. A multimodal biometric system based on finger vein images is proposed in this paper, by combining information from the index, middle and ring fingers of the hand. The essential characteristics from the finger vein images are extracted using a Convolutional Neural Network with a ReLU activation function. The input test image features are then compared with the features stored in the database using the correlation- based matching technique, and the match scores are fused using the arithmetic mean-based score level fusion. The performance of the proposed work is analyzed using the finger vein images from STUMULA -HMT database. The results reveal that the suggested multimodal biometric system outperformed the existing techniques, with a maximum accuracy of 99.83%.