Article
Details
Citation
Hancock PJB, Bruce V & Burton AM (1998) A comparison of two computer-based face identification systems with human perceptions of faces. Vision Research, 38 (15-16), pp. 2277-2288. https://doi.org/10.1016/S0042-6989%2897%2900439-2
Abstract
The performance of two different computer systems for representing faces was compared with human ratings of similarity and distinctiveness, and human memory performance, on a specific set of face images. The systems compared were a graphmatching system (e.g. Lades et al., 1993) and coding based on Principal Components Analysis (PCA) of image pixels (e.g. Turk & Pentland, 1991). Replicating other work, the PCA-based system produced very much better performance at recognising faces, and higher correlations with human performance with the same images, when the images were initially standardised using a morphing procedure and separate analysis of "shape" and "shape-free" components then combined. Both the graph-matching and (shape + shape-free) PCA systems were equally able to recognise faces shown with changed expressions, both provided reasonable correlations with human ratings and memory data, and there were also correlations between the facial similarities recorded by each of the computer models. However, comparisons with human similarity ratings of faces with and without the hair visible, and prediction of memory performance with and without alteration in face expressions, suggested that the graph-matching system was better at capturing aspects of the appearance of the face, while the PCA-based system seemed better at capturing aspects of the appearance of specific images of faces.
Keywords
PCA; similarity; distinctiveness; face recognition; Face perception; Facial recognition (Psychology) Graphic methods; Face recognition Computer simulation Paired comparisons
Journal
Vision Research: Volume 38, Issue 15-16
Status | Published |
---|---|
Publication date | 31/12/1998 |
Publication date online | 27/04/1999 |
URL | http://hdl.handle.net/1893/285 |
Publisher | Elsevier |
ISSN | 0042-6989 |
People (1)
People
Professor, Psychology