Ecole d'ingénieur et centre de recherche en Sciences du numérique

EURECOM submission to the Albayzin 2016 speaker diarization evaluation

Patino, Jose; Delgado, Hector; Evans, Nicholas; Anguera, Xavier

IBERSPEECH 2016, November 23-25, 2016, Lisbon, Portugal

Best Albayzin Evaluation System Award - Speaker Diarization evaluation (ex-aecquo)

This paper describes the speaker diarization system submitted by EURECOM for the Albayzin 2016 speaker diarization evaluation. This evaluation consists of segmenting broadcast audio documents according to di erent speakers and attributing those segments to the speaker who uttered them, without any prior information about the speaker identities nor their number. EURECOM system is based on the binary key speaker modelling, an ecient and compact speech and speaker representation. The proposed system does not require external training data: the test data itself is used for estimating the resources needed. The system delivered a diarization error rate (DER) of 11.93% on the development set.

Document Bibtex

Titre:EURECOM submission to the Albayzin 2016 speaker diarization evaluation
Mots Clés:speaker diarization, constant Q transform, ICMC, binary key, bi- nary key background model
Type:Conférence
Langue:English
Ville:Lisbon
Pays:PORTUGAL
Date:
Département:Sécurité numérique
Eurecom ref:5104
Copyright: © Springer. Personal use of this material is permitted. The definitive version of this paper was published in IBERSPEECH 2016, November 23-25, 2016, Lisbon, Portugal and is available at :
Bibtex: @inproceedings{EURECOM+5104, year = {2016}, title = {{EURECOM} submission to the {A}lbayzin 2016 speaker diarization evaluation}, author = {{P}atino, {J}ose and {D}elgado, {H}ector and {E}vans, {N}icholas and {A}nguera, {X}avier}, booktitle = {{IBERSPEECH} 2016, {N}ovember 23-25, 2016, {L}isbon, {P}ortugal}, address = {{L}isbon, {PORTUGAL}}, month = {11}, url = {http://www.eurecom.fr/publication/5104} }
Voir aussi: