skip to main content

SciTech ConnectSciTech Connect

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

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