By Richard Gray
At first glance Erin and Audri Nelson look identical. They have the same chocolate coloured hair, they share dark, almond-shaped eyes and have a similar impish smile.
Since birth their family, friends and teachers have mixed them up. When their high school prom dates came to their house in a small town in east Texas, the two boys could not tell which twin they were there to collect.
Yet despite their similarities, certain features do mark them apart. “We feel that we are more different than others perceive us to be,” explains Erin. “There are subtle differences in our faces – Audri has a slightly rounder face, maybe because she smiles more. Our lips also have a slightly different contour, but it is only noticeable if we are side by side in a picture.”
Those subtle cues are often missed by facial recognition software, which uses computers to automatically detect distinctive features and match them to images held in vast databases. And it is for this reason that Erin and Audri are now helping computer scientists, such as Kevin Bowyer at the University of Notre Dame in Indiana, to put their facial recognition and other “biometric identification” techniques to the ultimate test.