Self-organization of a sound source localization robot by perceptual cycle

We have built a system that can acquire the ability for sound source localisation by self-organization through repetition of motion and sensing. The learning model consists of two modules - a visual estimation module and an auditory estimation module of neural networks. The visual estimation module learns by direct inverse modeling and the auditory estimation module uses the output from the visual estimation module in learning. We conducted an experiment using a robot and a sound source to investigate the validity of the proposed module. The experimental results demonstrate that sound source localization ability can be acquired without supervision.