Graduate School and Research Center in Digital Sciences

ODESSA/PLUMCOT at Albayzin multimodal diarization challenge 2018

Maurice, Benjamin; Bredin, Hervé; Yin, Ruiqing; Patino, Jose; Delgado, Hector; Barras, Claude; Evans, Nicholas; Guinaudeau, Camille

IBERSPEECH 2018, 21-23 November 2018, Barcelona, Spain

Best System Award Iberspeech 2018 Albayzin Challenges - Multimodal Diarization

This paper describes ODESSA and PLUMCOT submissions to Albayzin Multimodal Diarization Challenge 2018. Given a list of people to recognize (alongside image and short video samples of those people), the task consists in jointly answering the two questions "who speaks when?" and "who appears when?". Both consortia submitted 3 runs (1 primary and 2 contrastive) based on the same underlying monomodal neural technologies: neural speaker segmentation, neural speaker embeddings, neural face embeddings, and neural talking-face detection. Our submissions aim at showing that face clustering and recognition can (hopefully) help to improve speaker diarization.

Document Hal Bibtex

Title:ODESSA/PLUMCOT at Albayzin multimodal diarization challenge 2018
Keywords:multimodal speaker diarization, face clustering
Type:Conference
Language:English
City:Barcelona
Country:SPAIN
Date:
Department:Digital Security
Eurecom ref:5731
Copyright: © ISCA. Personal use of this material is permitted. The definitive version of this paper was published in IBERSPEECH 2018, 21-23 November 2018, Barcelona, Spain and is available at :
Bibtex: @inproceedings{EURECOM+5731, year = {2018}, title = {{ODESSA}/{PLUMCOT} at {A}lbayzin multimodal diarization challenge 2018}, author = {{M}aurice, {B}enjamin and {B}redin, {H}erv{\'e} and {Y}in, {R}uiqing and {P}atino, {J}ose and {D}elgado, {H}ector and {B}arras, {C}laude and {E}vans, {N}icholas and {G}uinaudeau, {C}amille}, booktitle = {{IBERSPEECH} 2018, 21-23 {N}ovember 2018, {B}arcelona, {S}pain}, address = {{B}arcelona, {SPAIN}}, month = {11}, url = {http://www.eurecom.fr/publication/5731} }
See also: