A program capable of understanding false facial expressions was developed by University of Bradford researchers.
By analyzing the movement of a smile on a person’s face, the software can determine whether the expression is authentic or not. The most important movements that the software detects occur around the eyes; thus supporting the popular perception that the real smile appears from the eyes.
“Smiling is probably the most common facial expression and a powerful way to show positive emotions,”
says Hasan Ugail, a professor of Visual Computing at the university who led the research.
“Techniques of analyzing human facial expressions have evolved a lot in recent years, but distinguishing an authentic smile from a fake one remains difficult because people are not good at paying attention to it.”
The software first maps a person’s face through video, and locates the mouth, cheeks and eyes. Then it measures how these features move as the smile progresses and calculates the differences in motion between videos showing real and fake smiles.
The researchers tested the program on two different datasets, one with images of people with authentic smiles and one with images that had fake / ‘rigged’ smiles. As they found, there were significant differences in the way the mouths and cheeks were moved; however, the most pronounced deviations / fluctuations were in the eyes, with genuine smiles producing at least 10% more movement in the muscles that were there.
Concerning the utility of the program, Ugail says that it could contribute to the development of improved methods of interaction between humans and computers; e.g. better biometric systems. It could also be useful to scientists conducting social and medical research.