Muecas: a multi-sensor robotic head for affective human robot interaction and imitation

Area of Science:

• Robotics and Human-Computer Interaction
• Artificial Intelligence and Machine Learning
• Biomedical Engineering and Affective Computing

Background:

• Human-robot interaction (HRI) requires robots to understand and express human-like emotions and expressions.
• Existing robotic systems often lack the nuanced capabilities for naturalistic social interaction.
• The Muecas robotic head is designed to bridge this gap by integrating advanced perception and expression mechanisms.

Purpose of the Study:

• To present the design and development of Muecas, a multi-sensor humanoid robotic head.
• To detail the control systems and software architecture for real-time facial expression recognition and synthesis.
• To demonstrate the robotic head's capability for human emotion imitation and goal-based interaction.

Main Methods:

• Development of a 12-degrees-of-freedom humanoid robotic head with human-like configuration (eyes, eyebrows, mouth, neck).
• Implementation of direct control via the Facial Action Coding System (FACS) for expression synthesis.
• Creation of a software architecture using the RoboComp framework for real-time facial expression detection, recognition, classification, and generation.

Main Results:

• Successful design and construction of the Muecas robotic head by IADeX and RoboLab.
• Detailed kinematic and controller designs are provided.
• Experimental results demonstrate real-time recognition and imitation of human facial expressions, showcasing system flexibility and reliability.

Conclusions:

• The Muecas robotic head offers a robust platform for advanced human-robot interaction.
• Direct FACS control and the RoboComp framework facilitate third-party integration and development of imitation/goal-based systems.
• The system effectively enables robots to perceive, recognize, and generate human facial expressions in real time.