DEVELOPMENT OF MULTIMODAL BIOMETRIC SYSTEM FOR VOTERS’ VERIFICATION IN AN ELECTIONEERING PROCESS

Abstract

Electioneering process in most developing countries such as Nigeria normally is accompanied with election frauds such as double registration, ballot stuffing and other forms of irregularities. To solve this problem, in the 2015 Nigerian general election, electronic biometric registration and verification systems were introduced. It has been observed that in some cases on the day of election, some of the genuinely registered voters were not recognized by the biometric machine. Furthermore, some eligible voters who are physically challenged were also denied the opportunities of exercising their franchise because the system used for the registration made use of fingerprint only thereby denying those without hands or fingers to vote. A Multi-modal biometric system that combines Iris and Fingerprint is designed to checkmate most of these drawbacks. The designed system exhibits a two-tier structure. The first is the voters’ registration system which focuses on the enrolment, enhancement, feature extraction and storage of features. The second structure is the verification phase; this is a phase that compares the templates in the database system with the presented features on the Election Day. The test set of images used in this research work were gathered from 50 subjects drawn from a sample of 60 volunteers from Akoko South West Local Government Area in Ondo State. The volunteers comprised mostly of adults (<2% were above age 50) and both sexes were well represented (102 male, 58 female). The algorithm used to automatically segment the iris region from an eye image was Daugman’s Integro-Differential Operator. Iris features were extracted using 2D Gabor Filter which is a modified Gabor filters and the crossing Number (CN) concept was employed for minutiae extraction. The Hamming distance was the matching metric employed. The dual v-support Vector Machine (2ν-SVM) fusion algorithm was used to integrate the matching scores of the iris and fingerprint together with their corresponding weights. The system was developed to run on Windows and all systems with NET framework using Java programming language and MySQL. The recorded experimental values of False Acceptance Rate (FAR) 0.0%; False Rejection Rate (FRR) 0.05% and the average matching time of 0.1754s indicate that the proposed system is expected to create room for full participation of all eligible voters, eliminate multiple voting and also increase the confidence of the people in the electioneering process.