This chapter verses about the experimentation of the different proposed techniques in order to evaluate its suitability in the task of speaker diarization for meetings. It is done by first defining a baseline system to compare all algorithms to. Such baseline system is derived from the broadcast news mono-channel system with several improvements that were considered standard and necessary to the system as adapted to meetings.
Then a set of metrics used in the evaluation of the different techniques are described in detail. Next, the databases that are used to compare the algorithms performance with that of the baseline and the reference segmentations which are used in the experiments are explained and reasoned. Finally, the different experiments with the proposed algorithms are performed and results are explained.