Embedded and Integrated Audition for a Mobile Robot

AAAI Conferences 

"Aurally Informed Performance' for mobile robots operating in natural environments brings difficult challenges, such as: localizing sound sources all around the robot; tracking these sources as they or the robot move; separate the sources as a pre-processing step for recognition and processing, in real-time; dialogue management and interaction in crowded conditions; evaluating performances of the different processing components in open conditions. In this paper, we present how we address these challenges by describing our eight microphone system for sound source localization, tracking and separation, our ongoing work on its DSP implementation, and the use of the system on Spartacus, our mobile robot entry to AAAI Mobile Robot Competitions addressing human-robot interaction in open settings.