Download case study (PDF document)

Victims and witnesses of crime could soon be working with police to produce the next generation of E-FIT pictures of suspects thanks to groundbreaking work by OU researchers.
 
The project, undertaken with a number of police forces, involved designing a new computer system in which faces are generated by genetic algorithms, so that they gradually evolve based on choices made by the witness. This process means witnesses do not have to describe the face of the suspect or try and recall individual features – which, say the psychologists, could give a much more accurate approximation.

“The inherent problem with the old system is that the picture would be made up bit by bit,” said Prof Graham Pike. “The witness would be asked to describe the hair, and then choose through hundreds of pictures of hair until they got one that was close to what they had seen. Then they’d do the same with the eyes, the nose and so on. But we know as psychologists that we remember faces as a whole, not as made up of component parts.”

Prof Pike and colleagues Dr Nicola Brace and Dr Jim Turner, from the OU’s International Centre for Comparative Criminological Research, have worked with the police for 15 years examining how people recall faces and how they verbalise that memory. And their psychological expertise combined with new forensic imaging software created by co-investigators at the University of Kent has created a new, more powerful system.

“It’s very difficult for any of us to describe a face – even of someone we know very well – in a way that can be interpreted accurately,” he said. “With this new system, we show the witnesses a 3x3 grid of nine faces and ask them to pick the one that looks most like the person they remember. Choosing one face gives us a new screen of nine faces similar to that one, and we ask them again to choose the greatest likeness. That opens up a screen with another nine, and so it goes on until the witness indicates they have made the face as accurate as they can.

“And of course if they say something like ‘yes, he was just like that, but his nose was slightly bigger’, we can adapt the images.”

And the system is impressive, says Prof Pike, the Associate Dean of Research. “Psychological theory suggests that cutting out verbalization and the necessity to recall individual facial features makes the process much easier for the witness.”

It’s quicker, too – it’s said that the new system, currently being trialed by the Home Office, can give a more accurate approximation than a composite E-FIT in about 15 minutes – a quarter of the time taken under the old system.

“It is only a trial at the moment and the old system does still have one main advantage – an enormous database, created over decades, of E-FIT features and paraphernalia such as hats, glasses and moles,” said Prof Pike. “But of course our system is growing all the time.”

The project, part of a number of research strands titled “Working Well with Witnesses” , was funded by the Engineering and Physical Sciences Research Council. Visit the International Centre for Comparative Criminological Research (ICCCR) website for more information.