EUROSPEECH 2003 - INTERSPEECH 2003
8th European Conference on Speech Communication and Technology

Geneva, Switzerland
September 1-4, 2003

        

Multi-Speaker DOA Tracking Using Interactive Multiple Models and Probabilistic Data Association

Ilyas Potamitis, George Tremoulis, Nikos Fakotakis

University of Patras, Greece

The general problem addressed in this work is that of tracking the Direction of Arrival (DOA) of active moving speakers in the presence of background noise and moderate reverberation level in the acoustic field. In order to efficiently beamform each moving speaker on an extended basis we adapt the theory developed in the context of Multi-target Tracking for military and civilian applications to the context of microphone array. Our approach employs Wideband MUSIC and Interacting Multiple Model (IMM) estimators to estimate the DOAs of the speakers under different kinds of motion and sudden change in their course. Probabilistic Data Association (PDA) is used to disambiguate and resolve DOA measurements. The efficiency of the approach is illustrated on simulated and real room experiments dealing with the crossing trajectories of two speakers.

Full Paper

Bibliographic reference.  Potamitis, Ilyas / Tremoulis, George / Fakotakis, Nikos (2003): "Multi-speaker DOA tracking using interactive multiple models and probabilistic data association", In EUROSPEECH-2003, 517-520.