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

Utterance verification for text-dependent speaker recognition: a comparative assessment using the RedDots corpus

Kinnunen, Tomi; Sahidullah, Md; Kukanov, Ivan; Delgado, Héctor; Todisco, Massimiliano; Sarkar, Achintya; Thomsen, Nicolai; Hautamaki, Ville; Evans, Nicholas; Tan, Zheng-Hua

INTERSPEECH 2016, Annual Conference of the International Speech Communication Association, September 8-12, 2016, San Francisco, USA

Text-dependent automatic speaker verification naturally calls for the simultaneous verification of speaker identity and spoken content. These two tasks can be achieved with automatic speaker verification (ASV) and utterance verification (UV) technologies. While both have been addressed previously in the literature, a treatment of simultaneous speaker and utterance verification with a modern, standard database is so far lacking. This is despite the burgeoning demand for voice biometrics in a plethora of practical security applications. With the goal of improving overall verification performance, this paper reports different strategies for simultaneous ASV and UV in the context of short-duration, text-dependent speaker verification. Experiments performed on the recently released RedDots corpus are reported for three different ASV systems and four different UV systems. Results show that the combination of utterance verification with automatic speaker verification is (almost) universally beneficial with significant performance improvements being observed.

Document Bibtex

Titre:Utterance verification for text-dependent speaker recognition: a comparative assessment using the RedDots corpus
Mots Clés:Speaker recognition, Text-dependent, Utterance verification, Confidence measure
Type:Conférence
Langue:English
Ville:San Francisco
Pays:ÉTATS-UNIS
Date:
Département:Sécurité numérique
Eurecom ref:4934
Copyright: © ISCA. Personal use of this material is permitted. The definitive version of this paper was published in INTERSPEECH 2016, Annual Conference of the International Speech Communication Association, September 8-12, 2016, San Francisco, USA and is available at :
Bibtex: @inproceedings{EURECOM+4934, year = {2016}, title = {{U}tterance verification for text-dependent speaker recognition: a comparative assessment using the {R}ed{D}ots corpus}, author = {{K}innunen, {T}omi and {S}ahidullah, {M}d and {K}ukanov, {I}van and {D}elgado, {H}{\'e}ctor and {T}odisco, {M}assimiliano and {S}arkar, {A}chintya and {T}homsen, {N}icolai and {H}autamaki, {V}ille and {E}vans, {N}icholas and {T}an, {Z}heng-{H}ua}, booktitle = {{INTERSPEECH} 2016, {A}nnual {C}onference of the {I}nternational {S}peech {C}ommunication {A}ssociation, {S}eptember 8-12, 2016, {S}an {F}rancisco, {USA}}, address = {{S}an {F}rancisco, {\'{E}}{TATS}-{UNIS}}, month = {09}, url = {http://www.eurecom.fr/publication/4934} }
Voir aussi: