Face and Iris based Human Authentication using Deep Learning

Unique data protection is must in the modern world. Security for information technology and access to restricted areas, such as airports, governments, healthcare facilities, and military bases, are two areas where biometric systems using a person’s unique physiological or behavioral features have grown increasingly popular. In order to make a biometric system function reliably over a wide range of uses and settings, it is best to employ a multimodal biometric system that identifies individuals using a combination of factors. To further enhance the security, a multi-modal biometric authentication system could be employed to circumvent the constraints of a single-modal biometric authentication system. The suggested work describes a multimodal biometrics method that uses a Face and Iris (FI) based model. At first, multi-feature based programmed FI recognition in dim lighting is provided. The FI image is used as the input for the work proposed here. Methods based on normalization or lighting has been presented to obtain good performance under illumination variations. After FI images are preprocessed, the Gabor features, Local Binary Pattern (LBP) features, and phase congruency features are retrieved. Z score level fusion is used to combine the extracted features and generate a key. Deep Neural Network (DNN) classifiers are used to identify FI images based on the fused keys. The experimental outcomes prove the proposed system outperforms the current system. However, when applied to color FI images, it does not yield accurate recognition results.