Image & Video

Humanoid Robot Head with 21 DoF Scores 4.13/5 in Human-Likeness

New receptionist robot uses Llama and Whisper for natural conversations and re-identification.

Deep Dive

A team of researchers led by Tharusha Fonseka has developed an articulated humanoid robot head designed specifically for receptionist duties. The mechanical structure includes 21 degrees of freedom (DoF) covering the mouth, eyes, eyebrows, and neck, all covered with realistic silicone skin to mimic human appearance and expressions. The system runs a model-based architecture that integrates SCRFD (face detection), ArcFace (face recognition), and ByTetrack (re-identification) for vision, while Llama and Whisper handle natural language understanding and speech processing. Hardware support enables real-time operation, allowing the robot to recognize and re-identify individuals during interactions.

Quantitative evaluations measured conversational ability and re-identification accuracy, while a user study assessed emotional expressiveness and human-likeness. The head achieved an average human-likeness score of 4.13 out of 5, indicating strong realism and engagement. The work, accepted at the IEEE/ASME International Conference on Advanced Intelligent Mechatronics (AIM2026), addresses key limitations of existing humanoid heads such as high cost, mechanical complexity, and limited adaptability. By combining expressive mechanics with advanced AI, this platform brings human-robot interaction closer to natural, real-world use cases in service and hospitality environments.

Key Points
  • 21 degrees of freedom (DoF) for mouth, eyes, eyebrows, and neck with realistic silicone skin.
  • AI stack uses SCRFD, ArcFace, ByTetrack for face recognition and Llama + Whisper for NLP.
  • User study achieved 4.13/5 human-likeness score; accepted at IEEE/ASME AIM2026 conference.

Why It Matters

Affordable, expressive robot heads with real-time AI could transform reception and service roles in hospitality and healthcare.

📬 Get the top 10 AI stories daily