Tyler C W, Kontsevich L L, 2003, "Evolving facial expressions by reverse correlation with 2-D noise" Perception 32 ECVP Abstract Supplement
Evolving facial expressions by reverse correlation with 2-D noise
C W Tyler, L L Kontsevich
To study the ability of humans to read subtle changes in facial expression, we describe a novel application of the spatial reverse-correlation technique, consisting of adding samples of spatial noise to the image and categorising the results according to their effect on human perception of emotion. This added-noise method differs from the 'bubbles' technique of Gosselin and Schyns (2001 Vision Research 41 2261 - 2271) in permitting an evolutionary development to novel forms. An ambiguous facial expression (Leonardo's Mona Lisa) was taken as a base image, and different 50%-contrast binary noise samples were added on each trial. Twelve naïve observers judged the facial expression on each of one hundred trials and responded on a four-category scale from "very sad" to "very happy". The noise samples were then summed according to the response category, the sum added to the face, and the process iterated. Thus, the expression could be rapidly evolved to conform to the observer's implicit expression template. The expression seemed meaningful to the observers for each added-noise sample. The selected noise instances converged rapidly to a stable change of expression, making iterative noise cumulation an efficient method of evolving changes to explore the observers' expression space (given the very low likelihood of jumping to the final expression in one step). Selection of subregions of the cumulated noise revealed that the smile was carried entirely by the shape of the mouth region; the perception of smiling in the eyes was solely attributable to a configurational effect projecting from the mouth region. We conclude that behavioural reverse correlation with 2-D binary noise is an efficient means of exploring a complex pattern space such as that of facial expressions, and even of 'painting' novel visual imagery.
[Supported by NEI 13025.]
These web-based abstracts are provided for ease of seaching and access, but certain aspects (such as as mathematics) may not appear in their optimum form. For the final published version of this abstract, please see
ECVP 2003 Abstract Supplement (complete) size: 1712 Kb