A Naturalistic Multimodal Human Interaction Dataset: Annotated Visuo-Auditory Cue and Attention Data
This naturalistic multimodal dataset captures 27 everyday interaction scenarios with detailed visuo-spatial & auditory cue annotations and eye-tracking data, offering rich insights for research in human interaction, attention modeling, and cognitive media studies.
May 1, 2025
Modeling Visual Attention in Naturalistic Human Interactions
We explore how cues like gaze, speech, and motion guide visual attention in observing everyday interactions, revealing cross-modal patterns through eye-tracking data and structured event analysis.
January 30, 2025
Kinematic Primitives in Action Similarity Judgments: A Human-Centered Computational Model
This study compares human and computational judgments of action similarity, revealing that both rely heavily on kinematic features like velocity and spatial cues and that humans don't rely much on action semantics.
January 30, 2023
Attentional synchrony in films: A window to visuospatial characterization of events
This study analyzes how visuospatial features in film scenes shape attentional synchrony, using eye-tracking data to reveal how scene complexity guides visual attention during event perception.
September 22, 2022
Anticipatory Instances in Films: What Do They Tell Us About Event Understanding?
This study explores how visuospatial features in film scenes—like occlusion or movement—relate to anticipatory gaze and event segmentation, using eye-tracking data and multimodal analysis to uncover patterns in human event understanding.
June 1, 2022
Event Segmentation Through the Lens of Multimodal Interaction
This research develops a conceptual cognitive model to examine how multimodal cues—like gaze, motion, and speech—influence event segmentation and prediction in narrative media, using detailed scene analysis and eye-tracking data from naturalistic movie viewing.
September 17, 2021
Action similarity judgment based on kinematic primitives
This study compares human judgments and a kinematics-based computational model in recognizing action similarity, revealing that both rely heavily on low-level kinematic features, with the model showing higher sensitivity but also greater bias.
October 30, 2020