Map-generation and identification of multiple sound sources from robot in motion


The paper presents a multiple sound sources mapping system from a robot embedded microphone array. The robot localizes sound direction and recognizes what sound it is while the robot is in motion. Then the system estimates the positions of the sound sources using triangulation from a short time period of directional localization results. Three key components are denoted: 1) accurate directional localization and separation of multiple sound sources using a microphone array 2) separated sound recognition from a several tens of milliseconds input signal 3) sound position estimation using the RAndom SAmple Consensus (RANSAC) algorithm from a tracked sound stream. By combining these techniques, the proposed system provides surrounding sound information: “Where does the sound come from?” and “What is the sound?”. It works with short term signal input, and is helpful to initially notice surrounding events.

DOI: 10.1109/IROS.2010.5651131

15 Figures and Tables

Cite this paper

@article{Sasaki2010MapgenerationAI, title={Map-generation and identification of multiple sound sources from robot in motion}, author={Yoko Sasaki and Simon Thompson and Masahito Kaneyoshi and Satoshi Kagami}, journal={2010 IEEE/RSJ International Conference on Intelligent Robots and Systems}, year={2010}, pages={437-443} }