Low spatial frequency channels are more useful than high spatial frequency channels in classifying face emotional expressions, simulation of fMRI data

Abstract

Vuilleumier, Armony, Driver & Dolan (2003) have shown that amygdala responses to fearful expressions seem to be more activated by intact or low spatial frequency (LSF) faces than high spatial frequency (HSF) faces. The fMRI results suggest that LSF components processed by the magnocellular layers of the lateral geniculate nucleus (LGN) might be conveyed by a subcortical pathway activating the pulvinar, superior colliculus and finally the amygdala. This subcortical pathway was assumed to bypass the striate cortex in order to process LSF components faster than HSF components of visual stimuli. The purpose of the present study is to test the usefulness of LSF information as compared to HSF information and to original stimuli in a visual classification task performed by an artificial neural network. This model links a computational model of visual perception and a back-propagation classifier. The basic idea is i) to compress visual information by means of a perceptual model of vision and ii) to provide a distributed model of cognition with the above mentioned visual inputs. The results show that visual information conveyed by LSF faces, which is processed very fast by the human perceptual system, allows a distributed neural system to correctly categorize fearful or neutral faces. This is not the case for HSF components. These results suggest that high-speed connections from the magnocellular layers to the amygdala might be a fast and efficient way to perform classification of human faces with respect to their emotional expressions.

Footnotes

 Post-doctoral position funding by the Fyssen fundation