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

The I4U mega fusion and collaboration for NIST speaker recognition evaluation 2016

Lee, K A; Hautamäki, V; Kinnunen, T; Delgado, Hector; Todisco, Massimiliano; Evans, Nicholas; et al.

INTERSPEECH 2017, Annual Conference of the International Speech Communication Association, August 20-24, 2017, Stockholm, Sweden

The 2016 speaker recognition evaluation (SRE'16) is the latest edition in the series of benchmarking events conducted by the National Institute of Standards and Technology (NIST). I4U is a joint entry to SRE'16 as the result from the collaboration and active exchange of information among researchers from sixteen Institutes and Universities across 4 continents. The joint submission and several of its 32 sub-systems were among topperforming systems. A lot of efforts have been devoted to two major challenges, namely, unlabeled training data and dataset shift from Switchboard-Mixer to the new Call My Net dataset. This paper summarizes the lessons learned, presents our shared view from the sixteen research groups on recent advances, major paradigm shift, and common tool chain used in speaker recognition as we have witnessed in SRE'16. More importantly, we look into the intriguing question of fusing a large ensemble of subsystems and the potential benefit of large-scale collaboration.

Document Bibtex

Titre:The I4U mega fusion and collaboration for NIST speaker recognition evaluation 2016
Mots Clés:speaker recognition evaluation, fusion, benchmark, Call My Net
Type:Conférence
Langue:English
Ville:Stockholm
Pays:SUÈDE
Date:
Département:Sécurité numérique
Eurecom ref:5234
Copyright: © ISCA. Personal use of this material is permitted. The definitive version of this paper was published in INTERSPEECH 2017, Annual Conference of the International Speech Communication Association, August 20-24, 2017, Stockholm, Sweden and is available at :
Bibtex: @inproceedings{EURECOM+5234, year = {2017}, title = {{T}he {I}4{U} mega fusion and collaboration for {NIST} speaker recognition evaluation 2016}, author = {{L}ee, {K} {A} and {H}autam{\"a}ki, {V} and {K}innunen, {T} and {D}elgado, {H}ector and {T}odisco, {M}assimiliano and {E}vans, {N}icholas and et al. }, booktitle = {{INTERSPEECH} 2017, {A}nnual {C}onference of the {I}nternational {S}peech {C}ommunication {A}ssociation, {A}ugust 20-24, 2017, {S}tockholm, {S}weden}, address = {{S}tockholm, {SU}{\`{E}}{DE}}, month = {08}, url = {http://www.eurecom.fr/publication/5234} }
Voir aussi: