A MULTIMODAL APPROACH TO INITIALISATION FOR TOP-DOWN SPEAKER DIARIZATION OF TELEVISION SHOWS - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2010

A MULTIMODAL APPROACH TO INITIALISATION FOR TOP-DOWN SPEAKER DIARIZATION OF TELEVISION SHOWS

Simon Bozonnet
  • Fonction : Auteur
  • PersonId : 903988
Félicien Vallet
  • Fonction : Auteur
  • PersonId : 912779
Nicholas Evans
  • Fonction : Auteur
  • PersonId : 938450
Jean Carrive
  • Fonction : Auteur
  • PersonId : 844613

Résumé

This paper presents a new multimodal approach to speaker diarization of TV show data. We hypothesize that the intra-speaker variation in visual information might be less than that in the corresponding acoustic information and therefore might be better suited to the task of speaker model initialisa-tion. This is an acknowledged weakness of the computation-ally efficient top-down approach to speaker diarization that is used here. Experimental results show that a recently proposed approach to purification and the new multimodal approach to initialisation together deliver 22% and 17% relative improvements in diarization performance over the baseline system on independent development and evaluation datasets respectively.
Fichier non déposé

Dates et versions

hal-02747730 , version 1 (03-06-2020)

Identifiants

  • HAL Id : hal-02747730 , version 1

Citer

Simon Bozonnet, Félicien Vallet, Nicholas Evans, Slim Essid, Gael Richard, et al.. A MULTIMODAL APPROACH TO INITIALISATION FOR TOP-DOWN SPEAKER DIARIZATION OF TELEVISION SHOWS. Eusipco, 2010, aalborg, Denmark. ⟨hal-02747730⟩
33 Consultations
0 Téléchargements

Partager

Gmail Facebook X LinkedIn More