What happens when two suspects have the same DNA, voice and face as each other? How do you tell who’s guilty and who’s innocent? Which twin do you trust?
Blaming your identical twin for your crimes may be a thing of the past due to a new computer program that is able to pick out small differences in the facial features of identical twins.
Telling identical twins apart via a computer program may seem like an odd venture, but New Scientist explains how this computer program came to be, and a recent case in Canada proves just how valuable this kind of technology is.
In a case where two identical twins broke into a house together, issues arose when it came to figuring out who held the weapon and who wore the mask. Their identical DNA made it difficult to decipher who was involved in which aspects of the case. The DNA evidence pointed to both brothers and the court decided to charge them equally.
Currently, science is able to tell the DNA of twins apart; however, the tests to do so are very complex and very expensive. In an effort to find a more practical solution to identifying differences in Twin DNA, the US Federal Bureau of Investigations funds a project at the Twins Days Festival in Twinsburg, Ohio.
One researcher who attended the Festival, Marios Savvides of Carnegie Mellon University collected photos of twins at the event and found that while twins may be identical, they often leave traces of their individuality on their faces. Savvides found that twins making the same expression still have asymmetry within their faces.
In addition to finding these asymmetries, the researchers also looked at the change over time of the twins faces. As people grow and new habits form there are signs left behind such as how a person smiles or frowns, if they are a smoker or non smoker, weight lose and weight gain all affect the appearance of their face and more importantly the lines on the face.
Savvides and his team have taken this knowledge of asymmetries in twins and began developing a computer program that can both read twins faces and tell them apart. Using an algorithm they previously developed to tell the age of someone in a photo, they are able to read age lines and differences in facial line between twins. The program was tested on 638 identical twins with a success rate of 90%. Meaning that 9 times out of 10 the computer program could distinguish who was who between the twins.
This research is important because more and more crimes are being caught on tape. If one twin is committing a crime, this computer program could mean the difference between wrongly convicting an innocent individual and letting a criminal go free.
Kevin Boyer of the University of Notre Dame, who assisted with the research points out that this work is helpful in distinguishing individuals as well as twins. In the ever growing use of security cameras it will be more and more important to make sure who we think we see is in fact who committed the crime. The same techniques and computer software that are being used to tell twins apart can then be adapted and used for all faces.