A hybrid GMM speaker verification system for mobile devices in variable environments

Authors:
Tsang Ing Ren;George D. C. Cavalcanti;Dimas Gabriel;Hector N. B. Pinheiro
Affiliations:
Center of Informatics, Federal University of Pernambuco, Brazil;Center of Informatics, Federal University of Pernambuco, Brazil;Center of Informatics, Federal University of Pernambuco, Brazil;Center of Informatics, Federal University of Pernambuco, Brazil
Venue:
IDEAL'12 Proceedings of the 13th international conference on Intelligent Data Engineering and Automated Learning
Year:
2012

Citing 4
Cited 0

Algorithms for clustering data

Algorithms for clustering data
Connectionist ideas and algorithms

Communications of the ACM
Pattern Classification (2nd Edition)

Pattern Classification (2nd Edition)
Type-2 fuzzy Gaussian mixture models

Pattern Recognition

Quantified Score

Hi-index	0.00

Visualization

Abstract

This paper proposes a hybrid GMM speaker verification system for mobile devices in variable environments. A neural network based on backprobagation learning and a kNN algorithm were used to determine the probabilities in the GMM models to verify the speaker from a voice signal. A Voice Activity Detection (VDA) algorithm was also used to improve the voice identification ratio. The proposed method was tested using a database obtained from several different mobile devices in used in different environments, showing the robustness of the system.