Reza Safabakhsh

Papers

Total Citations

H-Index

About

Reza Safabakhsh is a researcher whose work sits at the intersection of human-robot interaction, computer vision, and natural language processing. His primary research areas include person recognition, biometrics, and automated video description. A key contribution is his work on non-intrusive person recognition for domestic service robots, where he pioneered methods that combine face and body information—including soft biometrics like body shape and clothing—to enable reliable identification even when a person’s face is not visible. His 2015 paper on this topic has garnered 5 citations, and he later improved the approach through weight adaptation of soft biometrics (3 citations). Safabakhsh also developed AUT-Talk, a Farsi talking head system that integrates text-to-speech with facial animation, demonstrating his versatility in multimodal systems. Most recently, his 2024 work on multi-sentence description of complex manipulation action videos pushes the boundaries of automated video understanding, aiming to generate varied levels of detail—a capability essential for assistive robotics and human-robot communication. Through these contributions, Safabakhsh has advanced the robustness of robotic perception and interaction in real-world, unconstrained environments.