Criminal identification given a human face

Police in the UK are set to test a new software technology that identifies crime suspects by ‘breeding’ their faces together and creating more realistic images than today’s photofit system.

Crime suspects could be identified by witnesses using technology that ‘breeds’ faces together, in UK police tests to be held later this year.

The software, the brainchild of Dr Chris Solomon and his team at the University of Kent, presents users with a dozen computer-generated, photo-realistic human faces based on gender, race and age details.

The faces chosen by a witness, or the victim of the crime as the closest match are then bred together to create a further set of faces that more accurately depict the suspect.

Solomon is working with Open University psychology lecturer Graham Pike and Kent police to develop and test the system.

The technology will create more realistic images than today’s photofit system, where a suspect’s face is built up from a range of eyes, noses, chins and other facial parts, said Solomon.

‘Part of the problem with current composite pictures is that they don’t really look human. We have a tendency to see faces globally. We don’t remember them just as a collection of individual features.’

The technique is combined with questions prepared by psychologists to help jog people’s memories, to ensure accurate facial recollection.

Solomon has been working on face recognition technology for five years, and has spent the past 18 months developing the system.

The Kent team has been awarded £105,000 from the EPSRC, plus a £45,000 DTI smart award. The Open University team has also received funding from the EPSRC.

Solomon has had talks with the chief constable of Kent, Sir David Philips, about testing the technology. The trials are expected to be later this year, when police will test it on witnesses to actual crimes.