Researchers explore the best biometrics for twins
22 July 2016 14:50 GMT

The subtle physical differences between “identical twins” has made them a bellwether of biometric technologies since the latter's inception, and the test case is often thrust onto new solutions and modalities. Now a team of researchers has set out to establish which technology can best tell twins apart.

Kevin W. Bowyer and Patrick J. Flynn from the Department of Computer Science and Engineering, University of Notre Dame have released a survey on the Biometric Identification of Identical Twins that explores the potential benefits and pitfalls of key technologies including face recognition, iris recognition, fingerprint matching and even voice biometrics.

The survey, which pulls together the literature to date in the area, identifies available datasets for research and points out topics of uncertainty, also suggests some ideas for possible future research.

Crucially, Bowyer and Flynn state that there is one major need to advance research in this area is a large, readily available multi-modal dataset that includes both identical and fraternal twins.

Among the pair’s findings, they state that iris recognition has the most unusual results for distinguishing between identical twins: “There is agreement that, with respect to the industry-standard, [John] Daugman-style approach to iris matching, MZ [monozygotic or identical twins’] iris codes are as different as those of unrelated persons. This is a stronger statement than can be made about other modalities. At the same time, there is also evidence that human observers can classify pairs of iris images as (a) belonging to twins or (b) belonging to unrelated persons at accuracy far above chance.”

On fingerprints, the experts state that there appears to be disagreement in the literature on the degree of similarity between MZ versus fraternal twins’ fingerprints.

"Srihari et al. state that '[t]he similarity of fingerprints of MZ twins is the same as the similarity between fingerprints of fraternal twins” [34], even though they and others find that fraternal twins have a much lower similarity in class of ridge pattern than do MZ twins.

The study concludes by stating that all of the studies on twins to date involve relatively small numbers of pairs of indentical twins, less than 50 pairs. Almost no attention has been paid specifically to fraternal twins. They also note that the effect of the twins’ age on the ability to distinguish between them has not been investigated. 

Image by Mikael Häggström :Häggström, Mikael. "Medical gallery of Mikael Häggström 2014". Wikiversity Journal of Medicine 1 (2). DOI:10.15347/wjm/2014.008. ISSN 20018762. - Own work, CC0, https://commons.wikimedia.org/w/index.php?curid=33081803