Multiple events tracking in sound tracks

Detecting and tracking broad sound classes in audio doc-uments is an important step toward structuration. In the case of complex audio scenes, such as TV broadcast sound tracks, one problem is that several audio events may oc-cur simultaneously. In this paper, we propose a two-step approach to detect superimposed events. The first step is a blind segmentation step, followed by an event detection step on each segment. In order to better evaluate the qual-ity of the system, new performance measures have been in-troduced, more suited to the superimposed events detection task. We also extend the two-step approach with an equiva-lent Viterbi-based event detection approach. 1