Multimodal Database: Biometric Authentication for Unconstrained Samples

Abstract

Biometrics provides a reliable authentication of a human in a wide variety of applications such as security systems, surveillance and human-computer interaction. Biometric system was started with utilization of a single biometric feature referring as a unimodal biometric system, which is unable to fulfill the security needs extensively in a highly sensitive environment and hence multibiometrics has emerged gaining more importance in the research area. Though there is a shortage of publicly available multimodal databases acquired in real unconstrained environment, a multimodal biometric system can succeed with the assistance of suitable multiple sensors providing higher accuracy rate than that of unimodal biometrics, of course subject to cost, time and subject’s acceptance This paper presents a new multimodal dataset which is developed using simple acquisition setup and devices to capture features belonging to the same person in uncontrolled scenarios. The dataset is composed of color images collected from 100 subjects (50 male and 50 female) under the age group 18–22. Totally 6 samples per trait were collected at different time internals between 2011 and 2014 with various occlusions. The dataset is also tested and analyzed by our developed biometric recognition system.