SPEAKER DIARIZATION: COMBINATION OF THE LIUM AND IRIT SYSTEMS

In this paper we present a new diarization system based on the combination of LIUM and IRIT systems. It contains speech detection followed by a GLR-BIC speaker segmenta-tion. Then we apply a BIC clustering followed by a CLR clustering. Moreover, we make some improvements by op-timizing clustering thresholds and by purifying the BIC clus-tering using Feature F0. Results show that our new system outperforms the best one of the ESTER evaluation campaign and that it is suitable for Broadcast News and meeting. 1