Student Presenter(s): Callyn Villanueva
Faculty Mentor: Houwei Cao
Department: Computer Science
School/College: College Engineering and Computing Sciences, New York City

The study of emotion recognition undoubtedly has an exciting future. Numerous researchers have made advances in the field; from analyzing facial expressions, body gestures and extracting speech features in hopes to improve the process of identifying human emotion. In this study, we explore subjects gaze patterns when identifying six universal emotions on videos of expressive talking faces in efforts to apprehend the impact of audio-visual channels on gaze behavior and perception. Stimuli for this experiment consists of carefully selected actor videos and audio from the Crowd-sourced Emotional Multimodal Actors dataset. The experiment consisted of two sessions (first session was sequential with 36 trials and second session was randomized with 36 trials) making a total of 72 trials. The participants' eye movements were recorded by using the Tobii Pro Nano eye-tracking system. We defined a set of area of-interest (AOI) regions consisting of 7 AOIs of general face areas and 15 AOIs related to specific Action Units (AUs) involved in the coding of the six basic emotions. Our findings show that participants overall response accuracy to an emotional facial expression was 79.2%. The random session had an accuracy of 80.5%, which was higher than sequential (77.8%). We also ran ANOVA tests to see differences in fixation time of AOI in different groups. We have found significant differences of a subset of AOI's across 6 emotions as well as Positive, Negative, and Neutral groups. We also see significances in the fixation time of AOI's in Congruent, Incongruent Natural, and Incongruent Synthetic groups.