Abstract
Perception of facial expressions has been primarily related to changes in the activation of the facial muscles. However, everyday experience suggests that neutral faces of different individuals are perceived as if they had different expressions. It is our hypothesis that the static configuration of facial components biases the perception of the face. Specifically, in this present work, we hypothesize that increasing the distance between the relative position of the nose, mouth, and eyes in neutral faces will result in the perception of sadness. Similarly, a decrease in those distances will result in the perception of increased anger. To test this hypothesis, twelve face images (768 × 576 pixels) with neutral expressions were selected. Each face was subjected to six morphing operations in which the distances among the eyes, nose, and mouth were varied. The eyebrows were moved in conjunction with the eyes, with the distance between the eyes and brows remaining constant at its initial value. The morph configurations included an increase (decrease) of 5, 10, 15, and 20 pixels. These stimuli were used in an experiment with one session for the angry class of images and another for the sad class. Pairs of images were sequentially presented for a period of 600 msec each, with a masked inter-stimulus interval of 500 msec. Subjects were asked to respond by keystroke to indicate if the second image was more, same or less sad (or angry) than the first image. Each image was presented randomly at one of the four morph levels or as the original image. Results demonstrated an increase of sadness (anger) perception as the distance between the eyes and mouth increased (decreased). A face-space based on the principal components of the shape of each face is shown to be sufficient for representing the variability in anger / sadness perception.
Supported in part by a grant from NIH.