The Gaze Dialogue Model: Nonverbal Communication in HHI and HRI

Area of Science:

• Cognitive Science
• Human-Robot Interaction
• Computational Neuroscience

Background:

• Human interaction relies on eye gaze for both motor control and communication.
• Gaze movements are crucial for understanding intentions and signaling actions during collaboration.
• Existing models often overlook the dynamic interplay of gaze in dyadic interactions.

Purpose of the Study:

• To develop a computational model, the Gaze Dialogue, that captures the interplay of eye movements during collaborative tasks.
• To correlate gaze fixation points with specific action types (individual vs. action-in-interaction).
• To infer human action types based on observed eye-gaze behavior.

Main Methods:

• Recording eye-gaze data from two participants engaged in collaborative tasks.
• Building a computational model encoding the dynamics of eye movements in dyadic interaction.
• Validating the model by using one participant's gaze data to predict the other's actions.

Main Results:

• The Gaze Dialogue model successfully encodes the interplay of eye movements during collaborative tasks.
• The model demonstrates the ability to correlate gaze fixation patterns with distinct action types.
• Validation confirmed the model's accuracy in inferring action types from gaze cues.

Conclusions:

• The Gaze Dialogue model provides a framework for understanding gaze-based communication in human collaboration.
• This model can be applied to enhance human-robot interaction by enabling robots to interpret and respond to human gaze.
• The developed humanoid robot controller effectively uses gaze cues for interpersonal coordination.