AES New York 2013
Tutorial T7
Friday, October 18, 10:30 am — 12:00 pm (Room 1E14)
T7 - Automatic Speaker Recognition, Voice Biometrics, and Forensic Speaker Comparison: Putting it All Together
Presenters:Catalin Grigoras, University of Colorado at Denver - Denver, CO, USA
Jeff M. Smith, National Center for Media Forensics - Denver, CO, USA; University of Colorado Denver - Denver, CO, USA
Abstract:
Automatic speaker recognition systems allow for an assessment of the similarities between two voices given a population of speakers in order to inform investigations or to grant biometric access to secure material. However, no internationally accepted standard for voice comparison procedure is in place for its use in forensics.
This tutorial will present the latest developments in automatic speaker recognition and considerations for its application in forensic speaker comparison. A methodology made up of two parts will be presented that relies on both automatic speaker recognition and semi-automatic vowel formant analysis. This two-pronged method allows the analyst to cross-validate the assessment of a speaker’s identity as a likelihood ratio. Additional discussion will include real cases from the past and very recent events.
Tutorial for all levels from those interested in an introduction to those with advanced knowledge in speech signal processing.
 | This session is presented in association with the AES Technical Committee on Audio Forensics |
