Comparison of scoring methods used in speaker recognition with Joint Factor Analysis

The aim of this paper is to compare different log-likelihood scoring methods, that different sites used in the latest state-of-the-art joint factor analysis (JFA) speaker recognition systems. The algorithms use various assumptions and have been derived from various approximations of the objective fu...

Full description

Saved in:
Bibliographic Details
Main Authors: Glembek, O., Burget, L., Dehak, N., Brummer, N., Kenny, P.
Format: Conference Proceeding
Language:English
Subjects:
Africa
Covariance matrix
Distributed computing
fast scoring
Gaussian distribution
GMM
Information analysis
Information technology
joint factor analysis
Performance loss
Speaker recognition
Speech analysis
Testing
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:The aim of this paper is to compare different log-likelihood scoring methods, that different sites used in the latest state-of-the-art joint factor analysis (JFA) speaker recognition systems. The algorithms use various assumptions and have been derived from various approximations of the objective functions of JFA. We compare the techniques in terms of speed and performance. We show, that approximations of the true log-likelihood ratio (LLR) may lead to significant speedup without any loss in performance.
ISSN:1520-6149
2379-190X
DOI:10.1109/ICASSP.2009.4960519