Accurate estimation of the Direction of Arrival (DOA) of a sound
source is an important prerequisite for a wide range of acoustic sig-
nal processing applications. However, in enclosed environments,
early reflections and late reverberation often lead to localization er-
rors. Recent work demonstrated that improved robustness against
reverberation can be achieved by clustering only the DOAs from
direct-path bins in the short-term Fourier transform of a speech sig-
nal of several seconds duration from a static talker. Nevertheless, for
moving talkers, short blocks of at most several hundred milliseconds
are required to capture the spatio-temporal variation of the source
direction. Processing of short blocks of data in reverberant envi-
ronment can lead to clusters whose centroids correspond to spuri-
ous DOAs away from the source direction. We therefore propose in
this paper a novel multi-detection source tracking approach that es-
timates the smoothed trajectory of the source DOAs. Results for re-
alistic room simulations validate the proposed approach and demon-
strate significant improvements in estimation accuracy compared to
single-detection tracking.