%0 Conference Proceedings %T End-to-end speaker segmentation for overlap-aware resegmentation %+ Équipe Structuration, Analyse et MOdélisation de documents Vidéo et Audio (IRIT-SAMoVA) %+ Laboratoire d'Informatique de l'Université du Mans (LIUM) %A Bredin, Hervé %A Laurent, Antoine %Z HPC resources of IDRIS under the allocation AD011012177 made by GENCI (Grand Équipement National de Calcul Intensif) %< avec comité de lecture %B Interspeech 2021 %C Brno, Czech Republic %8 2021-08-30 %D 2021 %Z 2104.04045 %K speaker diarization %K speaker segmentation %K voice activity detection %K overlapped speech detection %K resegmentation %Z Computer Science [cs]/Artificial Intelligence [cs.AI] %Z Computer Science [cs]/Computation and Language [cs.CL] %Z Computer Science [cs]/Neural and Evolutionary Computing [cs.NE]Conference papers %X Speaker segmentation consists in partitioning a conversation between one or more speakers into speaker turns. Usually addressed as the late combination of three sub-tasks (voice activity detection, speaker change detection, and overlapped speech detection), we propose to train an end-to-end segmentation model that does it directly. Inspired by the original end-to-end neural speaker diarization approach (EEND), the task is modeled as a multi-label classification problem using permutation-invariant training. The main difference is that our model operates on short audio chunks (5 seconds) but at a much higher temporal resolution (every 16ms). Experiments on multiple speaker diarization datasets conclude that our model can be used with great success on both voice activity detection and overlapped speech detection. Our proposed model can also be used as a post-processing step, to detect and correctly assign overlapped speech regions. Relative diarization error rate improvement over the best considered baseline (VBx) reaches 17% on AMI, 13% on DIHARD 3, and 13% on VoxConverse. %G English %2 https://hal-univ-lemans.archives-ouvertes.fr/hal-03257524/document %2 https://hal-univ-lemans.archives-ouvertes.fr/hal-03257524/file/2104.04045.pdf %L hal-03257524 %U https://hal-univ-lemans.archives-ouvertes.fr/hal-03257524 %~ UNIV-TLSE2 %~ UNIV-TLSE3 %~ CNRS %~ UNIV-LEMANS %~ UT1-CAPITOLE %~ LIUM %~ LIUM-LST %~ IRIT %~ IRIT-SAMOVA %~ ANR %~ IRIT-SI %~ IRIT-CNRS %~ TOULOUSE-INP %~ UNIV-UT3 %~ UT3-INP %~ UT3-TOULOUSEINP