Speaker modeling technique based on regression class for speaker identification with sparse training

Authors:
Zhonghua Fu;Rongchun Zhao
Affiliations:
School of Computer Science, Northwestern, Polytechnical University, Xi'an, P.R China;School of Computer Science, Northwestern, Polytechnical University, Xi'an, P.R China
Venue:
SINOBIOMETRICS'04 Proceedings of the 5th Chinese conference on Advances in Biometric Person Authentication
Year:
2004

Citing 1
Cited 1

Speaker identification and verification using Gaussian mixture speaker models

Speech Communication

Robust bootstrapping of speaker models for unsupervised speaker indexing

MCAM'07 Proceedings of the 2007 international conference on Multimedia content analysis and mining

Quantified Score

Hi-index	0.00

Visualization

Abstract

Speaker modeling technique with sparse training data is an active branch of robust speaker recognition research This paper presents a novel modeling approach named Multi-EigenSpace modeling technique based on Regression Class (RC-MES), which integrates the common eigenspace technique and the regression class (RC) idea of Maximum Likelihood Linear Regression (MLLR) RC-MES not only solves the problem of prior knowledge limitation of Gaussian Mixture Models (GMM) but also remedies the shortcoming of common eigenspace that confuses speaker differences and phoneme differences The eigenvoice analysis in RC can provide better discrimination ability between different speakers The experimental results on speaker identification of 75 males show that, when enrolment data is sparse, RC-MES provides significant improvement over GMM, and the number of eigenvoices in RC-MES is fewer than that in common eigenspace.