New bilingual speech databases for audio diarization