主催: 一般社団法人 日本機械学会
会議名: ロボティクス・メカトロニクス 講演会2017
開催日: 2017/05/10 - 2017/05/13
This paper presents a method of simultaneous localization and mapping (SLAM) for estimating the positions of multiple sound sources and those of stationary robots and synchronizing microphone arrays attached to those robots. Since each robot with a microphone array can solely estimate the directions of sound sources, the two-dimensional sound positions can be estimated from the source directions estimated by multiple robots. In addition, sound mixtures can be separated accurately by regarding distributed microphone arrays as one big array. To perform these tasks, the robot positions and synchronization between microphone arrays are necessary. The proposed method estimates the posterior distribution of the positions and time offsets and conducts source separation simultaneously in a Bayesian manner, given the observed signals. We conducted experiments using three robots and four sound sources. When the two of the model parameters (robot positions, sound source positions, and time offsets) were fixed to the correct value, the other one was correctly estimated and the observed signals were separated precisely. However, when all of the parameters were estimated simultaneously, they cannot be estimated correctly because of many local optimal solutions of the posterior distribution.