Face recognition experts developed the EvoFIT software and a new interview technique, boosting arrest rates in the UK and internationally.

Impacts

  • EvoFIT software has been used by police forces across the UK since 2007, when it was first used following the assault of an 11-year-old girl in Blackpool. As the result of the facial composite that was produced, two witnesses identified the offender.
  • The software was then trialled in forces in Lancashire and Derbyshire, where it directly led to arrests in 23 per cent of cases.
  • After modifications were made, EvoFIT was further trialled in Devon and Cornwall, and in Romania, where it led to an arrest rate of 36 per cent.
  • Finally after further modifications, EvoFIT was trialled in Humberside where it led to an arrest rate of 60 per cent.
  • From 2009 EvoFIT has been available as a commercial product managed by UCLan Business Services Ltd. Since the business started, it has been used over 3,000 times across 20 police forces in the UK, Romania, the Republic of Ireland, the US and Israel.
  • Due to the success of incorporating animated caricatures into EvoFIT, they have also been included in two further commercial composite systems used by police across the country: PRO-fit and EFIT-V.
  • Dr Charlie Frowd has trained around 150 police personnel in nine police forces in the UK and US on how to use caricaturing for these systems, as well as how to use the H-Cl interview method. As a result police forces now use both the character-based interview (H-CI) and dynamic caricature as standard to get the highest rate of identification.

"A quality product with images that are so good that a potential offender handed himself in to the police for a serious offence." (Devon and Cornwall Police)

About the research

Traditionally, creating photofits of criminal offenders' faces relied on a detailed witness description of the face, followed by description of features such as the nose and eyes, but the success rate with this approach is very low. Research has shown people recall faces by focusing on the whole face, rather than on individual features.

Dr Charlie Frowd from the University of Central Lancashire worked with Professor Peter Hancock at the University of Stirling and Professor Vicki Bruce at Newcastle University to develop EvoFIT (Evolving Facial Imaging Technique), a software designed to more closely match how humans recognise faces. EvoFIT currently produces composites with a suspect identification rate of 60 per cent, compared to only around 10 per cent from traditional facial feature systems.

The EvoFIT software provides the witness with a number of faces containing random features such as for eyes, noses and mouths, and blends the selected closest-match faces to produce a new, more refined set – filtering the photofit through consecutive rounds until the witness feels the resemblance with the criminal is as close as possible.

Dr Frowd also developed the 'holistic-cognitive interview' that can be combined with EvoFIT to reconstruct a face, asking eyewitnesses about the character of the target face (eg, how attractive or distinctive it is). In addition, his research has shown that the public are much more likely to recognise and identify a composite face if the image is stretched or viewed from the side, or if features are progressively caricatured, by systematically exaggerating a person's ears, forehead or nose.

ESRC-funded research, led by Dr Charity Brown at the University of Leeds, confirmed the effectiveness of this approach. Findings showed that when the holistic-cognitive interview is used in conjunction with EvoFIT, with the face essentially built from the inside out, and the image is viewed side-on, the composite is correctly named with an average of 74 per cent – which is the highest percentage recorded. This information has proved invaluable for police investigations and has been included in police training.