A Hybrid Multimodal Biometric Recognition System (HMBRS) based on Fusion of Iris, Face, and Finger Vein Traits

Biometric Recognition Systems (BRS) play a significant role in various applications to protect the data from attackers. Many traditional biometric systems rely on a single trait, such as the iris, face, or finger vein, and frequently run into challenges like environmental fluctuations and spoofing vulnerability. This study proposes a hybrid multimodal biometric recognition system (HMBRS) combining complex features of iris, face, and finger vein traits using deep learning techniques to overcome these challenges. The HMBRS model combines adaptive Convolutional Neural Networks (A-CNN) and Residual Networks (ResNets) as training models that help to extract the essential patterns from the given input images. Transfer learning is one domain that helps transfer the exciting patterns and textures in all the input images to the A-CNN, which allows for accurate BRS. The results show that the HMBRS achieves higher recognition accuracy, robustness against spoofing attacks, and resilience to variations in illumination and pose. The proposed approach obtained the performance for the Iris dataset with the accuracy-95.67, specificity-96.23, recall-97.45, and 97.34. For the face dataset with the accuracy-94.78, specificity-97.45, recall-96.12, and 96.21. For the finger vein traits dataset, the accuracy was 95.18, the specificity was 96.61, and the recall was 97.72 and 97.72.