Skip to main navigation Skip to search Skip to main content

Voice Doppelgänger Susceptibility among Racial and Gender Groups: IEEE CNS 21 Poster

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

1 Scopus citations

Abstract

Voice doppelganger describes a person or thing that sounds like another, which is considered a potential security risk for voice biometrics. Considering that racial or gender groups have different biological vocal structures, it is possible that these subgroups have different vulnerabilities to the voice doppelganger and voice biometrics. This study investigates if racial and gender disparities exist in the security risk of the voice doppelganger towards the voice biometric. We used three different metrics to measure voice similarity: fundamental frequency, MFCCs, and pitch. The result was that persons within gender and racial subgroups do indeed sound more similar to each other, with racial subgroups displaying more similarity than gender.

Original languageEnglish
Title of host publication2021 IEEE Conference on Communications and Network Security, CNS 2021
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9781665444965
DOIs
StatePublished - 2021
Event2021 IEEE Conference on Communications and Network Security, CNS 2021 - Tempe, United States
Duration: Oct 4 2021Oct 6 2021

Publication series

Name2021 IEEE Conference on Communications and Network Security, CNS 2021
Volume2021-January

Conference

Conference2021 IEEE Conference on Communications and Network Security, CNS 2021
Country/TerritoryUnited States
CityTempe
Period10/4/2110/6/21

Keywords

  • fairness
  • voice biometrics
  • voice doppelganger

Fingerprint

Dive into the research topics of 'Voice Doppelgänger Susceptibility among Racial and Gender Groups: IEEE CNS 21 Poster'. Together they form a unique fingerprint.

Cite this