Indoor multiple sound source tracking using refined TDOA measurements

Multiple sound source tracking in a reverberant environment is studied using multiple microphone arrays. The tracking problem is formulated into a sequential state estimation framework, where source coordinates are the states, and the time differences of arrival (TDOAs) are the measurements. A particle filter based on the random finite set theory is used to approximate the posterior likelihood of source locations given the TDOAs. To get more accurate TDOAs in challenging environments and thus further improve the tracking performance, a new approach is adopted. Preliminary TDOAs are estimated at selected time-frequency (TF) data points, and a TDOA histogram is formed. The final TDOAs are obtained by determining the values corresponding to the peaks of the resulted TDOA histogram. Simulations are performed to show the effectiveness of the proposed method.