Reproduce “Imitation of human motion achieves natural head movements for humanoid robots in an active-speaker detection task” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Bosong Ding, Murat Kirtay, Giacomo Spigler
Researchers at Tilburg University developed a generative modeling approach to enable a Nao humanoid robot to produce human-like head movements for active-speaker detection. Their system achieved an 85% reduction in active speaker detection inference time and generated motions that human participants preferred over the robot's default movements in a preliminary study.
Key insights
- VAEs can effectively learn the underlying kinematics of human head movements, producing diverse and natural-looking trajectories that generalize to new target points.
- Significant reduction in ASD model inference time (85%) through targeted optimizations allows for robust real-time active speaker detection on resource-constrained robots.
- The integration of generative motion modeling with real-time ASD enables a humanoid robot to exhibit human-like, coordinated gaze shifts, which are perceived as more natural by humans than default robot motions.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.