Applied Vision Association Christmas Meeting, Leuven, Belgium 19–20 December 2013

McIntyre, Alex

Abstract

For about 15 years, we have been attempting to create a software system (EvoFIT) that produces an identifiable composite when constructed of an unfamiliar face after a long retention interval (as is the usual case in police investigations). EvoFIT has now been shown to be effective in police field trials and has been used in 20 police forces. More recently, some developments have been very effective on their own to improve composite quality—these include: (a) encouraging constructors to focus on the overall (global) properties of the face during the initial interview (Holistic CI, H-CI), (b) constructing internal features first (masking external features), and (c) presenting composites for naming using a perceptual image-stretch format. In the current project, we explored whether these techniques would produce even more identifiable composites when combined. Forty participants looked at a video of an unfamiliar target face and, 24 hours later, received an interview to obtain a description of the face (either CI/ H-CI) and then constructed a composite using EvoFIT (either blurry externals/internals-only). Further participants were presented with the resulting composites to name, first front-on (the usual way to view the face) and then by looking at the face sideways (perceptual stretch). The three techniques were effective independently, replicating past research, but together produced composites with an effect that was more than just additive—to yield a mean naming level of 74% correct. A composite system using all three of these techniques would therefore appear to be effective for police-work.

Citation

McIntyre, A. (2014). Applied Vision Association Christmas Meeting, Leuven, Belgium 19–20 December 2013. Perception, 43(10), 1114-1134. https://doi.org/10.1068/ava13xm

• Journal Article Type: Article
• Acceptance Date: Jan 1, 2014
• Publication Date: 2014-10
• Deposit Date: Apr 11, 2019
• Publicly Available Date: Aug 20, 2019
• Journal: Perception
• Print ISSN: 0301-0066
• Electronic ISSN: 1468-4233
• Publisher: SAGE Publications
• Peer Reviewed: Peer Reviewed
• Volume: 43
• Issue: 10
• Pages: 1114-1134
• DOI: https://doi.org/10.1068/ava13xm
• Public URL: http://researchrepository.napier.ac.uk/Output/680382
• Contract Date: Aug 20, 2019