Research shows that a photo of a smiling person makes it easier to identify them than if they were putting on a neutral face. Could this help with AI facial recognition?
Humans are certainly the smiliest of animals.
While some animals, especially higher primates like chimpanzees, display grin-like expressions, they flash their teeth for different reasons than humans do.
Evolution played a role in the development of smiling as a subtle means of communication, but culture also has something to do with it.
Smiling is a human facial expression that looks effortless – even babies start to spontaneously smile long before they learn to utter sounds. It’s a feature that all humans can identify and relate to.
Now, new research suggests that people are better at identifying a smiling face than a neutral one.
As well as raising some sociological questions about this ability, it may also provide new methods of developing facial recognition AI.
Crack a Smile to Help Others ID you
Across all cultures, smiling is universally interpreted as a manifestation of affection, friendliness, and sociality.
Now, research from the University of York says that the more open a smile is in a photograph, the more likely the person will be recognized.
York researchers have already shown in past experiments that it’s hard for most people to match a pair of unfamiliar faces in photographs. This could mean serious concerns when identity fraud is brought into the equation.
Dr. Mila Mileva at York’s Department of Psychology wanted to investigate how smiles could help people identify unfamiliar individuals in different photos.
The research team conducted three studies where they asked participants to match photos of unfamiliar neutral faces with faces displaying full, toothy smiles.
When the images belonged to the same person, participants had a 9 percent improvement in their identification task. With photos of individuals who are different but look similar, their performance improvement dropped to 7 percent.
In another experiment, participants were asked to compare neutral facial expressions with faces having closed-lipped smiles. Then, in the final experiment, they had to look at the lower part of faces only to match photos.
The research definitively showed that big, toothy smiles are the best way of helping people to identify faces. Although the reason for this is still unclear, it could be answered with the help of facial recognition AI.
Machines Could use a “Smile” too
In the light of the present study’s findings, and as security at passport checkpoints go, there are three options for utilizing this tech for security or identification purposes.
Researchers suggested two options, either having people to crack a broad grin in their passport photo, or making smiling photos available to agents at checkpoints so they can improve their rates of identifying people.
However, a true practical solution would be facial recognition AI that takes smiles into account to improve their abilities at recognizing individuals.
Although still unclear, Machine Learning systems could possibly perform better than they do now if fed on smiling pictures.
Now we have to wait for other researchers to pick up on the work of York psychologists and see how AI would do in a similar experiment.
One of the biggest abilities of Deep Neural Networks is recognizing patterns, and smiles are a distinctive recurring pattern that could pose a solution for a number of issues in the development of facial recognition AI.
Smiles look all the same as they involve the same muscular mechanism and similar psychological grounds, yet they differ from person to person.
Everyone’s smile is different, and this could provide facial recognition AI with a resource that could certainly help it improve its abilities and accuracy over time.