Feature Level Fusion of Face and Fingerprint Biometric Traits for Universality

In recent days, biometric authentication systems have emerged as a popular method for identifying people based on their physiological and behavioral traits. Since most of the biometric systems only use one modality, it may cause issues such as noise in sensed data, intra-class differences, and similarities between classes, etc. In order to avoid some problems i.e., universality and security purposes multi-modal biometrics are proposed in this work. Multi-modal biometrics uses more than one modality for recognition. Various fusion techniques, including feature level, score level, decision level, etc., will be used to carry out these biometric recognition systems. In this study, we will put into practice a multi-modal biometric face and fingerprint identification system employing feature level fusion. The stages involved in building this system are, firstly it details the preprocessing techniques for both face and fingerprint and then the VGG16 model is built to carry out the further process i.e., extracting the individual features and then the individual model's features are classified using the support vector classifier SVC and got the accuracy of 90% for face and 93% for fingerprint. Next to do the fusion individual features are combined together using concatenation technique. Finally, the combined features are classified using the same classifier which is used for individual models and got the fused accuracy as 97%. This experimental result is carried out on the (Chinese academy of science’s institute of automation) CASIAFaceV5 and CASIA-FingerprintV5 datasets.