Wednesday, June 7, 2017

Cracking the Code of Face Recognition

Scientists at Caltech have come to a better understanding of how the brain recognize faces. They already knew from brain imaging studies that certain brain regions are especially active during a face recognition task.

But what those regions do and how they interact with each other has remained a mystery. One recent theory had suggested that individual cells were associated with specific faces. But that idea may not explain how the faces were coded in the first place.

The outcome of this study, which used monkeys as subjects, suggests that as few as 200 localized neural areas specialize in specific aspects or vectors of the face, such as the spacing between the eyes or the height of the forehead.

Using thousands of computer-generated faces that could be constructed from such vectors, scientists showed monkeys a set of individual human faces. Then, using solely the electrical signals transmitted from the monkeys' brains, their colleagues were able to generate a predicted face based only on that abstract information. The scientists were surprised how close the predicted face matched the actual face.

The findings still need to be replicated, but they promise to help in the development of artificial facial recognition technology. They also have relevance for portrait artists. According to the New York Times, "the brain's face cells respond to the dimensions and features of a face in an elegantly simple, though abstract, way." Perhaps that would explain why caricatures so elegantly define an individual.

New York Times
Scientific American 
Full text of scientific paper

1 comment:

essay best said...

Good post, so much information about face recognition, you did some really good job here, thanks for sharing it and keep posting more such posts