Paper Info
Title
A framework for speaker retrieval and identification through unsupervised learning
Abstract
•A speaker recognition approach based on unsupervised learning for retrieval and identification tasks.•The unsupervised learning algorithms employ a rank-based formulation, which can be applied to different features and modeling techniques.•Validation of the framework through MFCC and PLP features; VQ and GMM modeling; and RL-Sim and ReckNN unsupervised algorithms.•An experimental evaluation conducted on 3 public datasets, considering different speaker recognition tasks.•Effectiveness gains up to +56% on retrieval measures obtained by the unsupervised learning approach.
Year
DOI
Venue
2019
10.1016/j.csl.2019.04.004
Computer Speech & Language
Keywords
Field
DocType
Speaker recognition,Speaker retrieval,Unsupervised learning,Vector quantization,Gaussian mixture model,i-vector
Speaker identification,Ranking,Computer science,Speech recognition,Exploit,Vector quantization,Speaker recognition,Unsupervised learning,Labeled data,Mixture model
Journal
Volume
ISSN
Citations 
58
0885-2308
0
PageRank 
References 
Authors
0.34
0
2
Name
Order
Citations
PageRank
Victor de Abreu Campos100.34
Daniel Carlos Guimarães Pedronette230425.47