DOE PAGES title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: Reducing computation in an i-vector speaker recognition system using a tree-structured universal background model

Abstract

The majority of state-of-the-art speaker recognition systems (SR) utilize speaker models that are derived from an adapted universal background model (UBM) in the form of a Gaussian mixture model (GMM). This is true for GMM supervector systems, joint factor analysis systems, and most recently i-vector systems. In all of the identified systems, the posterior probabilities and sufficient statistics calculations represent a computational bottleneck in both enrollment and testing. We propose a multi-layered hash system, employing a tree-structured GMM–UBM which uses Runnalls’ Gaussian mixture reduction technique, in order to reduce the number of these calculations. Moreover, with this tree-structured hash, we can trade-off reduction in computation with a corresponding degradation of equal error rate (EER). As an example, we also reduce this computation by a factor of 15× while incurring less than 10% relative degradation of EER (or 0.3% absolute EER) when evaluated with NIST 2010 speaker recognition evaluation (SRE) telephone data.

Authors:
 [1];  [2]
  1. Sandia National Lab. (SNL-NM), Albuquerque, NM (United States)
  2. New Mexico State Univ., Las Cruces, NM (United States)
Publication Date:
Research Org.:
Sandia National Lab. (SNL-NM), Albuquerque, NM (United States)
Sponsoring Org.:
USDOE National Nuclear Security Administration (NNSA)
OSTI Identifier:
1140965
Report Number(s):
SAND-2014-2055J
Journal ID: ISSN 0167-6393; PII: S0167639314000582
Grant/Contract Number:  
AC04-94AL85000
Resource Type:
Accepted Manuscript
Journal Name:
Speech Communication
Additional Journal Information:
Journal Volume: 66; Journal Issue: C; Journal ID: ISSN 0167-6393
Publisher:
Elsevier
Country of Publication:
United States
Language:
English
Subject:
42 ENGINEERING; speaker recognition; clustering methods; tree graph

Citation Formats

McClanahan, Richard, and De Leon, Phillip L. Reducing computation in an i-vector speaker recognition system using a tree-structured universal background model. United States: N. p., 2014. Web. doi:10.1016/j.specom.2014.07.003.
McClanahan, Richard, & De Leon, Phillip L. Reducing computation in an i-vector speaker recognition system using a tree-structured universal background model. United States. https://doi.org/10.1016/j.specom.2014.07.003
McClanahan, Richard, and De Leon, Phillip L. Wed . "Reducing computation in an i-vector speaker recognition system using a tree-structured universal background model". United States. https://doi.org/10.1016/j.specom.2014.07.003. https://www.osti.gov/servlets/purl/1140965.
@article{osti_1140965,
title = {Reducing computation in an i-vector speaker recognition system using a tree-structured universal background model},
author = {McClanahan, Richard and De Leon, Phillip L.},
abstractNote = {The majority of state-of-the-art speaker recognition systems (SR) utilize speaker models that are derived from an adapted universal background model (UBM) in the form of a Gaussian mixture model (GMM). This is true for GMM supervector systems, joint factor analysis systems, and most recently i-vector systems. In all of the identified systems, the posterior probabilities and sufficient statistics calculations represent a computational bottleneck in both enrollment and testing. We propose a multi-layered hash system, employing a tree-structured GMM–UBM which uses Runnalls’ Gaussian mixture reduction technique, in order to reduce the number of these calculations. Moreover, with this tree-structured hash, we can trade-off reduction in computation with a corresponding degradation of equal error rate (EER). As an example, we also reduce this computation by a factor of 15× while incurring less than 10% relative degradation of EER (or 0.3% absolute EER) when evaluated with NIST 2010 speaker recognition evaluation (SRE) telephone data.},
doi = {10.1016/j.specom.2014.07.003},
journal = {Speech Communication},
number = C,
volume = 66,
place = {United States},
year = {Wed Aug 20 00:00:00 EDT 2014},
month = {Wed Aug 20 00:00:00 EDT 2014}
}

Journal Article:
Free Publicly Available Full Text
Publisher's Version of Record

Citation Metrics:
Cited by: 6 works
Citation information provided by
Web of Science

Save / Share: