This paper describes an open source software system for robot audition called HARK (Honda Research Institute Japan Audition for Robots with Kyoto University). HARK consists of a lot of modules including multi-channel audio input, sound source localization, sound source tracking, sound source separation and recognition of separated speech for robot audition based on the data-flow oriented software programming environment, FlowDesigner. By combining these modules using a GUI environment, a user can easily build a robot audition system for various types of robots and acoustic environments. Through HARK applications to Honda ASIMO and Robovie with different microphone settings, we showed high software portability and reusability of HARK.
