Introducing a new multimodal database from twins' biometric traits
Hamid Behravan, Karim Faez · 2013
This paper presents a new multimodal database from twins' biometric traits intended for twins and person authentication from multiple cues. The database consists of six unimodal biometric traits, namely two dimensional (2D) face images, fingerprints, offline handwritten texts, videos of moving faces, spectral and thermal face images. A total of 104 subjects corresponding to 52 pairs comprises the database from which 20 pairs are identical and the rest are fraternal twins. Besides biometric traits, personality traits and psychological characteristics of twins were also collected using two popular psychology questionnaires, Craig's Locus of Control (LOC) and the Big Five. Additionally, we conduct an experiment to measure human capability in distinguishing between identical twins. The result shows that untrained humans could classify identical twins with 82% accuracy using facial information and 76% accuracy using writing styles.