A Study of the Computational Space of Facial Expressions of Emotion
Ohio State University, Columbus OH
Investigators
Linked publications, trials & patents
Abstract
Project Summary Researchers generally agree that human emotions correspond to the execution of a number of computations by the nervous system. But, there is strong disagreement on what these computations are. One highly contentious point is the perception of emotion through facial expressions. That is, which are the emotion signals produced by a sender that are visually recognized by an observer? The overarching goal of our research is to identify these signals and specify the form and dimensions of the computational model of their visual recognition. Our general hypothesis is that the human visual system solves the inverse problem of production. In the first five years of funding, we have studied the hypothesis that consistent and differential facial muscle actions (called, action unites or AUs) are a subset of these computations executed by the nervous system. That is, when experiencing the same emotion, identical AUs are used by all people. Additionally, these AU activations are differential between emotions. Thus, our generally hypothesis states that the goal of the visual system is to identify which AUs are present in an image of a face. To date, we have completed several computational, behavioral and imaging studies favoring this hypothesis. These studies identified a previously unknown set emotive signals called compound facial expressions of emotion. However, these results were given by an analysis of posed facial expressions of emotion filmed in controlled conditions in the laboratory. The first specific aim of this renewal is to assess the validity of these results on facial expressions seen outside the laboratory (called ?in the wild?). Since the heterogeneity of spontaneous facial expressions observed outside of the lab is larger than those seen in posed expressions, we hypothesize that facial expressions in the wild communicate an even larger number of emotions than those previously identified. Also, these results study the computations executed by the nervous system that yield consistent and differential movements of one?s facial muscles only. In our second aim, we hypothesize that the computations executed by the nervous system also involve changes in blood flow, e.g., by increasing blood flow in the face when expressing anger. Specifically, we hypothesize that these changes are consistent within and differential between emotions. Thus, our general hypothesis implies that the visual system must identify these facial color changes. We will also test the alternative hypothesis that facial color is used to communicate other emotive variables, e.g., valence and arousal. Our third and final aim will examine the neural mechanisms associated to the computations studied in the previous two aims. Our studies are timely, because a lack of understanding of the computations of emotion poses a critical barrier to progress in basic and clinical research. Specifically, researchers are divided on whether emotion categories identified in the laboratory also exist in the wild, there is a poor understanding of which emotion signals are communicated through facial expressions, and it is unclear which neural mechanisms are associated with these computations.
View original record on NIH RePORTER →