Robust Speaker Identification in a Meeting with Short Audio Segments

The paper proposes a speaker identification scheme for a meeting scenario, that is able to answer the question "is somebody currently talking?", if yes, "who is it?". The suggested system has been designed to identify during a meeting conversation the current speaker from a set of pre-trained speaker models. Experimental results on two databases show the robustness of the approach to the overlapping phenomena and the ability of the algorithm to correctly identify a speaker with short audio segments.