Design of a multibiometric system for user authentication

Abstract

The major objectives of this work are to create an effective multimodal biometric system that addresses all of the difficulties that currently exist in biometric systems. This study analysis several techniques to this end, and then propose a novel hybrid optimization process. First, the performance of Uni-modal biometric systems for both Iris and Fingerprint characteristics is compared in this work. So, the major work has been done on feature level fusion, and it has been determined that feature extraction is the most essential component in feature level fusion. It was discovered that texture features did not specify all of the sample properties, thus edge and key point features were included to the proposed work. When all of these characteristics are merged, it is discovered that the extracted features are in huge quantities, making managing them a challenging task. To cope with this, a new hybrid optimization is employed, and feature selection is added to this stage. This hybrid optimization is a combination of the firefly and IWD algorithms, both of which are fast and intelligent in nature and aid in extracting only the relevant features from the samples. Fusion is then performed using the sum algorithm, which is a robust fusion algorithm in which the sum of all the features calculated first, and then the mean value is stored in the database for further processing. Overall, the suggested system performs significantly better and meets all of the objectives. MATLAB is used as a simulator for implementation and performance evaluation. For both Iris and Fingerprint traits, the CASIA and IITD data sets were employed. The total number of samples in this data set is 200, with 100 coming from CASIA and the rest from IITD. Some forgeries samples are also utilized for testing. The performance of this proposed work is compared to that of all others levels with different techniques, and it is found that the proposed technique is outperforming than others. newline