Colloquium - Details

Kommunikationstechnisches Kolloquium: Acoustic scene mapping for human-robot interaction

Dr. Christine Evers
2. Dezember 2015
15:00 Uhr
Hörsaal 4G IKS

The ability to explore the surrounding environment is a major precondition for robot autonomy and is therefore crucial for Human-Robot Interaction. Audio signals obtained from microphone arrays integrated in the robot platform contain information about the environment that can be exploited constructively for scene analysis, especially in situations where visual sensors suffer from limited Field of View. Acoustic scene mapping is a challenging task as microphone arrays can often localize sound sources only in terms of their instantaneous directions of arrival whilst missing source-sensor range. Furthermore, due to reverberation and speech inactivity, sound source localization in enclosed spaces is subject to missing detections and spurious clutter measurements. Source tracking approaches are therefore required in order to smooth the localization estimates, and apply temporal models of the source dynamics to infer trajectories of speaker positions from the localized source directions. This talk discusses recent developments in multi-speaker tracking and acoustic scene mapping for human-robot interaction. After a discussion of sound source localization approaches using pseudo-spherical robot microphone arrays we present an overview of tracking approaches in the literature. A new theoretical framework for multi-speaker tracking and acoustic scene mapping is introduced that provides improved robustness against the adverse affects of dominant reflections due to reverberation. The theoretical treatment is followed by the presentation and discussion of simulation results.

back