A multimodal biometric system for personal verification based on different level fusion of iris and face traits

Recognizing users who access technological resources is one of the required processes to secure these resources. Based on previous studies, the most robust user recognition method is the biometric systems as these systems use unique behavioral or physical data to recognize individuals. There are two types of biometric systems: unimodal and multimodal biometric systems. Unimodal systems use only one biometric trait for user recognition, while multimodal systems use multiple traits. Experimental studies have shown that unimodal biometric systems have some problems related to performance and accuracy. In order to avoid the limitations of unimodal systems, multimodal biometric systems are deployed. In this paper, we propose a multimodal biometric system for human verification using a deep learning algorithm. The system depends on the face and iris identification model. The model uses an end-to-end convolutional neural network (CNN) for extracting features and identifying the user without using any image detection techniques. In the proposed model, two different approaches were tested to fuse the two biometric traits: feature-level fusion and score-level fusion. The experiments showed that the proposed model achieved an accuracy rate of 99.22% using the feature-level fusion approach, and using the score-level fusion approach resulted in an accuracy rate of 100%.