Graduate School and Research Center in Digital Sciences

Further optimisations of constant Q cepstral processing for integrated utterance verification and text-dependent speaker verification

Delgado, Héctor. Todisco, Massimiliano. Sahidullah, Md. Sarkar, Achintya K.. Evans, Nicholas; Kinnunen, Tomi; Tan, Zheng-Hua

SLT 2016, IEEE Workshop on Spoken Language Technology, December 13-16, 2016, San Juan, Puerto Rico

Many authentication applications involving automatic speaker verification (ASV) demand robust performance using short-duration, fixed or prompted text utterances. Text constraints not only reduce the phone-mismatch between enrolment and test utterances, which generally leads to improved performance, but also provide an ancillary level of security. This can take the form of explicit utterance verification (UV). An integrated UV + ASV system should then verify access attempts which contain not just the expected speaker, but also the expected text content. This paper presents such a system and introduces new features which are used for both UV and ASV tasks. Based upon multi-resolution, spectro-temporal analysis and when fused with more traditional parameterisations, the new features not only generally outperform Mel-frequency cepstral coefficients, but also are shown to be complementary when fusing systems at score level. Finally, the joint operation of UV and ASV greatly decreases false acceptances for unmatched text trials.

Document Doi Bibtex

Title:Further optimisations of constant Q cepstral processing for integrated utterance verification and text-dependent speaker verification
Keywords:speaker verification, utterance verification, text dependent, constant Q transform
Type:Conference
Language:English
City:San Juan
Country:PUERTO RICO
Date:
Department:Digital Security
Eurecom ref:4990
Copyright: © 2016 IEEE. Personal use of this material is permitted. However, permission to reprint/republish this material for advertising or promotional purposes or for creating new collective works for resale or redistribution to servers or lists, or to reuse any copyrighted component of this work in other works must be obtained from the IEEE.
Bibtex: @inproceedings{EURECOM+4990, doi = {http://dx.doi.org/10.1109/SLT.2016.7846262}, year = {2016}, title = {{F}urther optimisations of constant {Q} cepstral processing for integrated utterance verification and text-dependent speaker verification}, author = {{D}elgado, {H}{\'e}ctor. {T}odisco, {M}assimiliano. {S}ahidullah, {M}d. {S}arkar, {A}chintya {K}.. {E}vans, {N}icholas and {K}innunen, {T}omi and {T}an, {Z}heng-{H}ua}, booktitle = {{SLT} 2016, {IEEE} {W}orkshop on {S}poken {L}anguage {T}echnology, {D}ecember 13-16, 2016, {S}an {J}uan, {P}uerto {R}ico}, address = {{S}an {J}uan, {PUERTO} {RICO}}, month = {12}, url = {http://www.eurecom.fr/publication/4990} }
See also: