Share
Blog
Cite
Print
Email
BookmarkScienceDaily (Sep. 14, 2007) — A new technique which enables us to identify faces more accurately could make security systems more reliable.
See also:
Humans are very good at identifying faces we know, said Dr Rob Jenkins, a psychologist at the University of Glasgow. He was delivering this year's Joseph Lister BA Award Lecture at the BA Festival of Science on September 11, 2007.
However, he said, we are much less accomplished when the face is unknown to us. This is almost always the case with the faces captured by CCTV cameras or used to check at a passport barrier.
In our burgeoning world of electronic communications, there are governmental plans to multiply CCTV cameras, electronic passport photos, and bring in identity cards. All in the name of "security".
But exactly how “secure” will we be if, in fact, we can't reliably identify the faces? One tragic example such a failure is the unnecessary killing of Jean Charles de Mendez during the July 2005 London terrorist attacks.
Our brain uses a highly specialised region to recognise familiar faces called the FFA (fusiform face area, an area of the brain nestled between the outer segment and the centre). Despite being able locate the FFA using modern imaging techniques, scientists still don't know how it works. Dr Jenkins has ingeniously side-stepped this problem.
He and his team have developed a new technique by "mimicking nature's solution" to help us out or, as he puts it, "extract the essence" of a face.
They use around 12 pictures of a single face in a variety of different environments and poses to find the "average" face. This means getting rid of anything distracting (light strength, directions) that "isn't useful information for our visual identity system", Dr Jenkins said.
From a technical view point, Dr Jenkins argues that his software is superior to that of the Home Office, which has been perfecting highly complex facial analysis programmes.
Amazingly, the "average" face has advantages over a "normal" face in almost every situation: correct matching is more frequent than for pictures or film.
Another remarkable feature of the "average" face is that it can help you recognise someone at any stage of their life. Using pictures of different ages actually makes the "average" better.
