Faye Skelton

Dr. Faye Skelton is a face recognition researcher and Lecturer in cognitive psychology at Edinburgh Napier University. Her research interests centre on forensic applications of face recognition, including eyewitness identification, police facial composites, and more recently, so-called “Super-Recognisers”.

Skelton obtained her PhD in face recognition from Lancaster University in 2004, focusing on how children learn new faces. After her PhD, whilst lecturing at the University of Central Lancashire, her research took a more applied angle, focusing on the improvement of police facial composite images in collaboration with Dr. Charlie Frowd. Examining more traditional feature-based (PROfit) and newer evolutionary (EvoFIT) composite software systems, she has explored ways of improving witness’ memory using interviews, changes to composite production procedures, and post-production techniques for enhancing the likeness of composites to targets. This work has dramatically improved correct naming rates of composites in laboratory tests and police field trials, ultimately helping the police to catch more offenders.

Facial composite images (often known as ‘E-Fits’) are visual likenesses of suspects based on a witness’ memory, and are typically produced using specialist software and trained operators. These images are frequently circulated via the media in order to generate lines of enquiry, but have historically suffered from very poor recognition and naming rates (~3%). Improvements have been made in the last decade primarily due to new software, which harnesses whole-face recognition rather than relying on detailed recall of individual facial features. Further improvements to composite accuracy have been made by making changes to the witness interview and composite production technique in order to improve the quality of the important ‘internal’ facial features, morphing composites from multiple witnesses, and altering the format of the image for publication. Recent research shows that combining some of these techniques can result in accurate naming rates in excess of 70%.