"TalkPrinting"

Improving speaker recognition by modeling stylistic features

Sachin Kajarekar, Mustafa (Kemal) Sonmez, Luciana Ferrer, Venkata Gadde, Anand Venkataraman, Elizabeth Shriberg, Andreas Stolcke, Harry Bratt

Research output: Contribution to journalArticle

Abstract

Automatic speaker recognition is an important technology for intelligence gathering, law enforcement, and audio mining. Conventional speaker recognition systems, which are based on independent short-term spectral samples, suffer from a lack of noise robustness and are unable to model a speaker's idiosyncratic stylistic features. This paper describes "TalkPrinting", a program of research aimed at adding such stylistic features to conventional systems. Results on three preliminary systems based on stylistic features demonstrate that (1) the new features alone carry significant speaker information; (2) they also carry significant complementary information compared to the conventional features; and (3) they provide increasing improvements in performance with increasing test durations.

Original languageEnglish (US)
Pages (from-to)350-354
Number of pages5
JournalLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume2665
StatePublished - 2003
Externally publishedYes

Fingerprint

Feature Modeling
Law Enforcement
Speaker Recognition
Law enforcement
Intelligence
Noise
Technology
Research
Noise Robustness
Mining
Demonstrate
Model

ASJC Scopus subject areas

  • Computer Science(all)
  • Biochemistry, Genetics and Molecular Biology(all)
  • Theoretical Computer Science

Cite this

"TalkPrinting" : Improving speaker recognition by modeling stylistic features. / Kajarekar, Sachin; Sonmez, Mustafa (Kemal); Ferrer, Luciana; Gadde, Venkata; Venkataraman, Anand; Shriberg, Elizabeth; Stolcke, Andreas; Bratt, Harry.

In: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), Vol. 2665, 2003, p. 350-354.

Research output: Contribution to journalArticle

Kajarekar, Sachin ; Sonmez, Mustafa (Kemal) ; Ferrer, Luciana ; Gadde, Venkata ; Venkataraman, Anand ; Shriberg, Elizabeth ; Stolcke, Andreas ; Bratt, Harry. / "TalkPrinting" : Improving speaker recognition by modeling stylistic features. In: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). 2003 ; Vol. 2665. pp. 350-354.
@article{c5db59b0ade84ad4a113f367faa9de2a,
title = "{"}TalkPrinting{"}: Improving speaker recognition by modeling stylistic features",
abstract = "Automatic speaker recognition is an important technology for intelligence gathering, law enforcement, and audio mining. Conventional speaker recognition systems, which are based on independent short-term spectral samples, suffer from a lack of noise robustness and are unable to model a speaker's idiosyncratic stylistic features. This paper describes {"}TalkPrinting{"}, a program of research aimed at adding such stylistic features to conventional systems. Results on three preliminary systems based on stylistic features demonstrate that (1) the new features alone carry significant speaker information; (2) they also carry significant complementary information compared to the conventional features; and (3) they provide increasing improvements in performance with increasing test durations.",
author = "Sachin Kajarekar and Sonmez, {Mustafa (Kemal)} and Luciana Ferrer and Venkata Gadde and Anand Venkataraman and Elizabeth Shriberg and Andreas Stolcke and Harry Bratt",
year = "2003",
language = "English (US)",
volume = "2665",
pages = "350--354",
journal = "Lecture Notes in Computer Science",
issn = "0302-9743",
publisher = "Springer Verlag",

}

TY - JOUR

T1 - "TalkPrinting"

T2 - Improving speaker recognition by modeling stylistic features

AU - Kajarekar, Sachin

AU - Sonmez, Mustafa (Kemal)

AU - Ferrer, Luciana

AU - Gadde, Venkata

AU - Venkataraman, Anand

AU - Shriberg, Elizabeth

AU - Stolcke, Andreas

AU - Bratt, Harry

PY - 2003

Y1 - 2003

N2 - Automatic speaker recognition is an important technology for intelligence gathering, law enforcement, and audio mining. Conventional speaker recognition systems, which are based on independent short-term spectral samples, suffer from a lack of noise robustness and are unable to model a speaker's idiosyncratic stylistic features. This paper describes "TalkPrinting", a program of research aimed at adding such stylistic features to conventional systems. Results on three preliminary systems based on stylistic features demonstrate that (1) the new features alone carry significant speaker information; (2) they also carry significant complementary information compared to the conventional features; and (3) they provide increasing improvements in performance with increasing test durations.

AB - Automatic speaker recognition is an important technology for intelligence gathering, law enforcement, and audio mining. Conventional speaker recognition systems, which are based on independent short-term spectral samples, suffer from a lack of noise robustness and are unable to model a speaker's idiosyncratic stylistic features. This paper describes "TalkPrinting", a program of research aimed at adding such stylistic features to conventional systems. Results on three preliminary systems based on stylistic features demonstrate that (1) the new features alone carry significant speaker information; (2) they also carry significant complementary information compared to the conventional features; and (3) they provide increasing improvements in performance with increasing test durations.

UR - http://www.scopus.com/inward/record.url?scp=35248812449&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=35248812449&partnerID=8YFLogxK

M3 - Article

VL - 2665

SP - 350

EP - 354

JO - Lecture Notes in Computer Science

JF - Lecture Notes in Computer Science

SN - 0302-9743

ER -