Fusing Face and Iris: A Deep/Machine Learning Approach for Advanced Biometric Recognition

In the digital era, the increasing demand for secure and reliable biometric systems has driven the shift towards multimodal fusion, such as combining face and iris recognition, to enhance accuracy and robustness. Traditional single-modality systems often struggle with challenges like variations in lighting, facial expressions, and noise, leading to higher error rates, including False Acceptance Rate (FAR) and False Rejection Rate (FRR). To address these limitations, we propose a novel machine learning-based score-level fusion method that integrates face and iris features. Using Local Binary Patterns (LBP) for face feature extraction and Gabor filters for iris, we generate robust feature vectors for each modality. These are then combined through a weighted sum approach, balancing contributions from both modalities to significantly improve recognition performance and reduce errors. This method effectively overcomes the limitations of traditional techniques by exploiting the complementary strengths of face and iris recognition. Our experiments, conducted on the CASIA Webface dataset and CASIA iris dataset with 5-fold cross-validation, demonstrated a notable improvement, achieving a 1% error rate, outperforming existing state-of-the-art biometric recognition systems.