Original title: Human-robot conversation with multiple participants in noisy public spaces
Divesh Lala, Yogeeswaran Muthukumaran, Vincent Fernandes, Kazushi Kato +9 · September 2026
Talking to robots in loud public spaces is tough because background noise drowns out human voices. To solve this, researchers designed a special audio setup that cleans up speech. It uses a single multi-channel microphone arrayA device with multiple microphones that captures sound from different directions to better separate voices., which is a collection of several microphones grouped together, to pick up sound from different directions. The system separates and enhances speech from several people at once. It also creates spatial audioSound processing that gives the listener a sense of where each voice is located in space., an effect that makes sound feel like it comes from specific places around you. The team tested their setup in two situations. One was with an android named ERICA. The other used mobile Teleco robots, where one robot served as an avatar for a remote human operator to enable more immersive multi-party conversations.
Sitting at a crowded dinner table, you try to follow a conversation but other voices and clinking dishes make it hard to pick out individual words. The difficulty is that all the sounds mix together, so you can't isolate who's saying what. In this paper, the audio system is like having a personal assistant at each seat who whispers only the relevant parts of the conversation directly to you, filtering out the rest. It allows the robot to 'hear' each speaker distinctly, even in a noisy group setting.
Robots and avatars are entering busy public spaces like expos, where noise makes group conversations hard to follow. This system helps them understand several people at once, making interactions more reliable for both autonomous robots and remote operators.
The abstract describes a proof-of-concept demonstrated at the 2025 World Expo. It does not provide details on performance metrics, scalability to larger crowds, or how it handles varying noise levels. It focuses on the design and demonstration rather than extensive testing or comparison with other systems.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
Human-robot conversation with multiple participants in noisy public spaces — Divesh Lala, Yogeeswaran Muthukumaran, Vincent Fernandes et al.
Abstract licence: CC0
Explanation generated by Synthesis from the abstract — it is not the original paper. This is a preprint, so it hasn't necessarily been peer reviewed.